Automated Machine Learning with Low Code - AMA

Name: Automated Machine Learning with Low Code - AMA | Open Video
Uploaded: 2023-11-06T19:58:45+00:00
Duration: 1 h 11 min 28 s

0:04
[Music]
0:46
[Music]
0:46
[Music] [Applause]
1:09
hi my name is aimy boys I'm a car
1:11
hi my name is aimy boys I'm a car
1:11
hi my name is aimy boys I'm a car developer advocate here at Microsoft
1:14
developer advocate here at Microsoft
1:14
developer advocate here at Microsoft [Music]
1:15
[Music]
1:15
[Music] [Applause]
1:18
[Applause]
1:18
[Applause] [Music]
1:23
[Music]
1:23
[Music] [Applause]
1:26
[Applause]
1:26
[Applause] [Music]
1:36
[Applause]
1:38
[Applause]
1:38
[Applause] [Music]
1:43
[Music]
1:43
[Music] [Applause]
1:44
[Applause]
1:44
[Music]
1:44
[Music] [Applause]
1:45
[Applause]
1:45
[Applause] [Music]
1:45
[Music]
1:45
[Music] [Applause]
1:51
[Music]
1:53
[Music]
1:53
[Music] [Applause]
2:09
[Music]
2:09
[Music]
2:09
[Music] [Applause]
2:17
hi everyone my name is Stephan Simon and
2:20
hi everyone my name is Stephan Simon and
2:20
hi everyone my name is Stephan Simon and welcome back to C sharp corner life show
2:22
welcome back to C sharp corner life show
2:22
welcome back to C sharp corner life show in this episode of ask me anything
2:24
in this episode of ask me anything
2:24
in this episode of ask me anything series we have Emma Kate joining us from
2:27
series we have Emma Kate joining us from
2:27
series we have Emma Kate joining us from United Kingdom who is the cloud Taylor
2:29
United Kingdom who is the cloud Taylor
2:29
United Kingdom who is the cloud Taylor for Advocate hi Amy how are you yeah
2:32
for Advocate hi Amy how are you yeah
2:32
for Advocate hi Amy how are you yeah very good thank you very good
2:34
very good thank you very good
2:34
very good thank you very good what an incredible introduction oh my
2:36
what an incredible introduction oh my
2:36
what an incredible introduction oh my goodness I'm gonna have to ask you if I
2:39
goodness I'm gonna have to ask you if I
2:39
goodness I'm gonna have to ask you if I can borrow that that was that so III
2:43
can borrow that that was that so III
2:43
can borrow that that was that so III didn't tell you I was always telling you
2:45
didn't tell you I was always telling you
2:45
didn't tell you I was always telling you the video followed by the countdown so I
2:49
the video followed by the countdown so I
2:49
the video followed by the countdown so I just wanted to keep it surprise so every
2:51
just wanted to keep it surprise so every
2:51
just wanted to keep it surprise so every why didn't you go ahead and tell who I
2:53
why didn't you go ahead and tell who I
2:53
why didn't you go ahead and tell who I mean if you had an intro video but still
2:55
mean if you had an intro video but still
2:55
mean if you had an intro video but still I think the audience would like to know
2:56
I think the audience would like to know
2:56
I think the audience would like to know right
2:57
right
2:57
right we have people joining us from across
2:59
we have people joining us from across
2:59
we have people joining us from across the globe sous chef on it's a global
3:01
the globe sous chef on it's a global
3:01
the globe sous chef on it's a global community and you are coming on a
3:03
community and you are coming on a
3:03
community and you are coming on a platform for the very first time so
3:05
platform for the very first time so
3:05
platform for the very first time so people would like to know who is Emma
3:06
people would like to know who is Emma
3:06
people would like to know who is Emma Kate what do where she lives so why
3:10
Kate what do where she lives so why
3:10
Kate what do where she lives so why don't you go and introduce yourself sure
3:12
don't you go and introduce yourself sure
3:12
don't you go and introduce yourself sure yeah so hi everyone and my name is Amy
3:15
yeah so hi everyone and my name is Amy
3:15
yeah so hi everyone and my name is Amy Boyd I am a cloud developer advocate at
3:20
Boyd I am a cloud developer advocate at
3:20
Boyd I am a cloud developer advocate at Microsoft and what does butts that
3:22
Microsoft and what does butts that
3:22
Microsoft and what does butts that actually mean because that's quite an
3:23
actually mean because that's quite an
3:23
actually mean because that's quite an unusual title it's only used in the tech
3:26
unusual title it's only used in the tech
3:26
unusual title it's only used in the tech industry Adams basically what I get to
3:29
industry Adams basically what I get to
3:29
industry Adams basically what I get to do is I work or I like to think of
3:32
do is I work or I like to think of
3:32
do is I work or I like to think of myself as sitting in the middle of
3:34
myself as sitting in the middle of
3:34
myself as sitting in the middle of community and developers on one side and
3:37
community and developers on one side and
3:37
community and developers on one side and then our internal engineering teams on
3:40
then our internal engineering teams on
3:40
then our internal engineering teams on the other and so one of the things I
3:42
the other and so one of the things I
3:42
the other and so one of the things I really enjoy doing is you like using our
3:45
really enjoy doing is you like using our
3:45
really enjoy doing is you like using our technology sharing news cases for our
3:48
technology sharing news cases for our
3:48
technology sharing news cases for our technology with our with any developers
3:51
technology with our with any developers
3:51
technology with our with any developers and data scientists and but then
3:53
and data scientists and but then
3:53
and data scientists and but then actually hearing from all of you people
3:56
actually hearing from all of you people
3:56
actually hearing from all of you people as well out there who are using our
3:57
as well out there who are using our
3:57
as well out there who are using our technologies and taking that feedback
3:59
technologies and taking that feedback
3:59
technologies and taking that feedback back to engineering so that we can make
4:03
back to engineering so that we can make
4:03
back to engineering so that we can make sure that our products are exactly what
4:05
sure that our products are exactly what
4:05
sure that our products are exactly what people are looking for as well as being
4:07
people are looking for as well as being
4:07
people are looking for as well as being hopefully exactly what you want to use
4:10
hopefully exactly what you want to use
4:10
hopefully exactly what you want to use for your to solve your problems that
4:12
for your to solve your problems that
4:12
for your to solve your problems that sounds really interesting job profile
4:14
sounds really interesting job profile
4:14
sounds really interesting job profile Amy I mean you work as a as a mediator
4:17
Amy I mean you work as a as a mediator
4:17
Amy I mean you work as a as a mediator between the product team and the coming
4:20
between the product team and the coming
4:20
between the product team and the coming you get an opportunity to work to
4:22
you get an opportunity to work to
4:22
you get an opportunity to work to understand what is very happening with
4:23
understand what is very happening with
4:23
understand what is very happening with the products and then you also get a
4:25
the products and then you also get a
4:25
the products and then you also get a feedback and what is happening with the
4:27
feedback and what is happening with the
4:27
feedback and what is happening with the community truly someone's very exciting
4:29
community truly someone's very exciting
4:29
community truly someone's very exciting we also know you have you have been in
4:30
we also know you have you have been in
4:30
we also know you have you have been in this Microsoft very very long time and
4:33
this Microsoft very very long time and
4:33
this Microsoft very very long time and you are known for your community stuff
4:35
you are known for your community stuff
4:35
you are known for your community stuff so is we have for you today so every
4:38
so is we have for you today so every
4:38
so is we have for you today so every what are we covering today I mean it
4:39
what are we covering today I mean it
4:39
what are we covering today I mean it does says automated machine learning
4:41
does says automated machine learning
4:41
does says automated machine learning with low code so what will be the basic
4:44
with low code so what will be the basic
4:44
with low code so what will be the basic agenda and what are the target audience
4:46
agenda and what are the target audience
4:46
agenda and what are the target audience that we're looking for today
4:47
that we're looking for today
4:47
that we're looking for today sure yeah so yeah that's there and title
4:52
sure yeah so yeah that's there and title
4:52
sure yeah so yeah that's there and title suggests we're going to talk about
4:53
suggests we're going to talk about
4:53
suggests we're going to talk about something called automated machine
4:55
something called automated machine
4:55
something called automated machine learning and it is something that is
4:57
learning and it is something that is
4:57
learning and it is something that is used across the industry so it's not
4:59
used across the industry so it's not
4:59
used across the industry so it's not just the Microsoft thing but what I'm
5:01
just the Microsoft thing but what I'm
5:01
just the Microsoft thing but what I'm gonna show you is in Microsoft Azure
5:04
gonna show you is in Microsoft Azure
5:04
gonna show you is in Microsoft Azure we're going to have a look at the
5:06
we're going to have a look at the
5:06
we're going to have a look at the machine learning service now don't worry
5:09
machine learning service now don't worry
5:09
machine learning service now don't worry if you've not used it before it's not a
5:12
if you've not used it before it's not a
5:12
if you've not used it before it's not a problem we're actually gonna focus on
5:14
problem we're actually gonna focus on
5:14
problem we're actually gonna focus on the low code element so we're only going
5:16
the low code element so we're only going
5:16
the low code element so we're only going to focus on UI elements as well as
5:19
to focus on UI elements as well as
5:19
to focus on UI elements as well as things like rest api calls and some
5:22
things like rest api calls and some
5:22
things like rest api calls and some jason code so really like don't worry
5:25
jason code so really like don't worry
5:25
jason code so really like don't worry we're not gonna be there is a whole
5:28
we're not gonna be there is a whole
5:28
we're not gonna be there is a whole obviously like python and our SDK
5:30
obviously like python and our SDK
5:30
obviously like python and our SDK available for this technology but I fell
5:33
available for this technology but I fell
5:33
available for this technology but I fell actually to introduce these concepts it
5:35
actually to introduce these concepts it
5:35
actually to introduce these concepts it would be easier to show you and walk you
5:37
would be easier to show you and walk you
5:37
would be easier to show you and walk you through it using the UI but that sounds
5:40
through it using the UI but that sounds
5:40
through it using the UI but that sounds very interesting
5:41
very interesting
5:41
very interesting Amy and I think you can go ahead and
5:44
Amy and I think you can go ahead and
5:44
Amy and I think you can go ahead and maybe if you want I can fingers screen
5:46
maybe if you want I can fingers screen
5:46
maybe if you want I can fingers screen into the stream I think low code is
5:48
into the stream I think low code is
5:48
into the stream I think low code is something that is really getting popular
5:50
something that is really getting popular
5:50
something that is really getting popular these days and it may be a power app it
5:53
these days and it may be a power app it
5:53
these days and it may be a power app it may be UI flows even Microsoft is
5:55
may be UI flows even Microsoft is
5:55
may be UI flows even Microsoft is committed to enhance their technology
5:58
committed to enhance their technology
5:58
committed to enhance their technology with the latest features so it's all you
6:00
with the latest features so it's all you
6:00
with the latest features so it's all you now I will be there in the life some
6:01
now I will be there in the life some
6:01
now I will be there in the life some people who will to watch it but it's all
6:03
people who will to watch it but it's all
6:03
people who will to watch it but it's all you guys can see your screen and there
6:05
you guys can see your screen and there
6:05
you guys can see your screen and there you go I love that today of course we
6:13
you go I love that today of course we
6:13
you go I love that today of course we have to have bit on the screen our
6:15
have to have bit on the screen our
6:15
have to have bit on the screen our mascot and we have a physical one just
6:19
mascot and we have a physical one just
6:19
mascot and we have a physical one just here as well he sits on my desk with me
6:22
here as well he sits on my desk with me
6:22
here as well he sits on my desk with me all the time but yeah sure we get
6:25
all the time but yeah sure we get
6:25
all the time but yeah sure we get started yeah sure it's all you now Amy
6:28
started yeah sure it's all you now Amy
6:28
started yeah sure it's all you now Amy perfect so yeah huge thank you everyone
6:31
perfect so yeah huge thank you everyone
6:31
perfect so yeah huge thank you everyone for joining us and with
6:33
for joining us and with
6:33
for joining us and with going to be talking a bit about
6:34
going to be talking a bit about
6:35
going to be talking a bit about automated machine learning and reducing
6:38
automated machine learning and reducing
6:38
automated machine learning and reducing that time to insight using low code so
6:42
that time to insight using low code so
6:42
that time to insight using low code so the the main driver here is to share
6:45
the the main driver here is to share
6:45
the the main driver here is to share with the developed community some of the
6:48
with the developed community some of the
6:48
with the developed community some of the great tooling and powerful machine
6:51
great tooling and powerful machine
6:51
great tooling and powerful machine learning offerings that are out there
6:53
learning offerings that are out there
6:53
learning offerings that are out there without having to have extensive machine
6:57
without having to have extensive machine
6:57
without having to have extensive machine learning knowledge Python code in our
6:59
learning knowledge Python code in our
6:59
learning knowledge Python code in our code and etc and there is a really nice
7:02
code and etc and there is a really nice
7:02
code and etc and there is a really nice spectrum I always think of all the
7:07
spectrum I always think of all the
7:07
spectrum I always think of all the different services we see so we can see
7:09
different services we see so we can see
7:09
different services we see so we can see something like what Stephen just
7:12
something like what Stephen just
7:12
something like what Stephen just mentioned there which was all the stuff
7:14
mentioned there which was all the stuff
7:14
mentioned there which was all the stuff around power apps and power automate and
7:17
around power apps and power automate and
7:17
around power apps and power automate and the AI builder in there is absolutely
7:20
the AI builder in there is absolutely
7:20
the AI builder in there is absolutely phenomenal it's a really amazing service
7:23
phenomenal it's a really amazing service
7:23
phenomenal it's a really amazing service so powerful and yet so easy to use then
7:26
so powerful and yet so easy to use then
7:26
so powerful and yet so easy to use then you might come down the spectrum a bit
7:28
you might come down the spectrum a bit
7:28
you might come down the spectrum a bit to something like the azure cognitive
7:29
to something like the azure cognitive
7:29
to something like the azure cognitive services quick REST API calls to get
7:33
services quick REST API calls to get
7:33
services quick REST API calls to get your machine learning knowledge back
7:35
your machine learning knowledge back
7:35
your machine learning knowledge back very very low amounts of sort of deep
7:39
very very low amounts of sort of deep
7:39
very very low amounts of sort of deep technical knowledge needed and then we
7:41
technical knowledge needed and then we
7:41
technical knowledge needed and then we start to run into the Azure machine
7:43
start to run into the Azure machine
7:43
start to run into the Azure machine learning space and in that space we have
7:45
learning space and in that space we have
7:45
learning space and in that space we have things like you might have heard of the
7:47
things like you might have heard of the
7:47
things like you might have heard of the designer or automated machine learning
7:49
designer or automated machine learning
7:49
designer or automated machine learning and then also for all the experts in the
7:52
and then also for all the experts in the
7:52
and then also for all the experts in the industry so thinking of your data
7:55
industry so thinking of your data
7:55
industry so thinking of your data scientist type persona people who code
7:57
scientist type persona people who code
7:57
scientist type persona people who code in Python and are they use things like
8:00
in Python and are they use things like
8:00
in Python and are they use things like Jupiter notebooks we have a lot around
8:02
Jupiter notebooks we have a lot around
8:02
Jupiter notebooks we have a lot around the SDKs that work alongside our Azure
8:06
the SDKs that work alongside our Azure
8:06
the SDKs that work alongside our Azure machine learning service to really make
8:08
machine learning service to really make
8:08
machine learning service to really make sure you have that enterprise and to end
8:11
sure you have that enterprise and to end
8:11
sure you have that enterprise and to end production ready machine learning
8:14
production ready machine learning
8:14
production ready machine learning service and so we're gonna sit right in
8:16
service and so we're gonna sit right in
8:16
service and so we're gonna sit right in the middle I am gonna talk some machine
8:19
the middle I am gonna talk some machine
8:19
the middle I am gonna talk some machine learning concepts however we will walk
8:22
learning concepts however we will walk
8:22
learning concepts however we will walk through that there is loads of amazing
8:24
through that there is loads of amazing
8:24
through that there is loads of amazing documentation I'll post some great links
8:26
documentation I'll post some great links
8:26
documentation I'll post some great links in the chat for you and so that you can
8:28
in the chat for you and so that you can
8:28
in the chat for you and so that you can continue your learning but I just really
8:30
continue your learning but I just really
8:30
continue your learning but I just really want to show you what's possible and
8:33
want to show you what's possible and
8:33
want to show you what's possible and give you an idea of the types of things
8:35
give you an idea of the types of things
8:35
give you an idea of the types of things that you can solve using this so um
8:39
that you can solve using this so um
8:39
that you can solve using this so um first things first rate what it was his
8:41
first things first rate what it was his
8:41
first things first rate what it was his automated machine learning because
8:43
automated machine learning because
8:43
automated machine learning because actually machine learning and I've
8:45
actually machine learning and I've
8:45
actually machine learning and I've always thought it quite automated
8:47
always thought it quite automated
8:47
always thought it quite automated and just like code if you write you know
8:50
and just like code if you write you know
8:50
and just like code if you write you know a few lines of code and then basically
8:52
a few lines of code and then basically
8:52
a few lines of code and then basically loop it you're going to be going through
8:55
loop it you're going to be going through
8:55
loop it you're going to be going through many many many pieces of data in order
8:58
many many many pieces of data in order
8:58
many many many pieces of data in order to train a machine learning model to
9:00
to train a machine learning model to
9:00
to train a machine learning model to understand your problem so machine
9:02
understand your problem so machine
9:02
understand your problem so machine learning now it's heart is very very
9:04
learning now it's heart is very very
9:04
learning now it's heart is very very automated however let me tell you not
9:07
automated however let me tell you not
9:07
automated however let me tell you not what automated machine learning is or
9:09
what automated machine learning is or
9:09
what automated machine learning is or but why it came about so in the machine
9:12
but why it came about so in the machine
9:12
but why it came about so in the machine learning space one of the things and
9:14
learning space one of the things and
9:14
learning space one of the things and that you find is that you're doing a lot
9:17
that you find is that you're doing a lot
9:17
that you find is that you're doing a lot of tweaking so even once you build a
9:21
of tweaking so even once you build a
9:21
of tweaking so even once you build a piece of code that does something for
9:23
piece of code that does something for
9:23
piece of code that does something for machine learning at you then it's ever
9:25
machine learning at you then it's ever
9:25
machine learning at you then it's ever really like the first thing that you
9:26
really like the first thing that you
9:26
really like the first thing that you write that's actually going to be useful
9:28
write that's actually going to be useful
9:28
write that's actually going to be useful you've really got to experiment with it
9:30
you've really got to experiment with it
9:30
you've really got to experiment with it so let me give you a scenario of what we
9:34
so let me give you a scenario of what we
9:34
so let me give you a scenario of what we do here so an example scenario is how
9:37
do here so an example scenario is how
9:37
do here so an example scenario is how much is this car worth and we have lots
9:40
much is this car worth and we have lots
9:40
much is this car worth and we have lots and lots of data based on lots of lots
9:42
and lots of data based on lots of lots
9:42
and lots of data based on lots of lots of different cars and what we want to do
9:45
of different cars and what we want to do
9:45
of different cars and what we want to do is we want to teach your machine
9:46
is we want to teach your machine
9:46
is we want to teach your machine learning model how to understand or give
9:50
learning model how to understand or give
9:50
learning model how to understand or give us some kind of probability of
9:52
us some kind of probability of
9:52
us some kind of probability of prediction of how much it thinks this
9:55
prediction of how much it thinks this
9:55
prediction of how much it thinks this specific car is worth and so one of the
9:58
specific car is worth and so one of the
9:58
specific car is worth and so one of the things that comes up when we start
10:00
things that comes up when we start
10:00
things that comes up when we start building is that actually model creation
10:02
building is that actually model creation
10:02
building is that actually model creation is very time consuming so when we write
10:06
is very time consuming so when we write
10:06
is very time consuming so when we write in this code we write one set and eat
10:09
in this code we write one set and eat
10:09
in this code we write one set and eat for each time we do we build one model
10:12
for each time we do we build one model
10:12
for each time we do we build one model the first thing we do is to say well
10:14
the first thing we do is to say well
10:14
the first thing we do is to say well which features so whether it's mileage
10:17
which features so whether it's mileage
10:17
which features so whether it's mileage car brand the the year the car was made
10:22
car brand the the year the car was made
10:22
car brand the the year the car was made and then we might need to look at okay
10:25
and then we might need to look at okay
10:25
and then we might need to look at okay which algorithm am I going to choose
10:27
which algorithm am I going to choose
10:27
which algorithm am I going to choose first and then each algorithm has
10:30
first and then each algorithm has
10:30
first and then each algorithm has different parameters so for example on
10:33
different parameters so for example on
10:33
different parameters so for example on my first try I might go right I'm going
10:35
my first try I might go right I'm going
10:35
my first try I might go right I'm going to pick these three features I'm going
10:38
to pick these three features I'm going
10:38
to pick these three features I'm going to choose this algorithm and then this
10:41
to choose this algorithm and then this
10:41
to choose this algorithm and then this algorithm has lots of different
10:43
algorithm has lots of different
10:43
algorithm has lots of different parameters that I need to set and then
10:46
parameters that I need to set and then
10:46
parameters that I need to set and then out of the end of that becomes a model
10:48
out of the end of that becomes a model
10:48
out of the end of that becomes a model and that model has some form of
10:51
and that model has some form of
10:52
and that model has some form of evaluation metric so this is a generic
10:54
evaluation metric so this is a generic
10:54
evaluation metric so this is a generic evaluation budget but it might be
10:56
evaluation budget but it might be
10:56
evaluation budget but it might be something like accuracy it might be
10:58
something like accuracy it might be
10:58
something like accuracy it might be something like
10:59
something like
10:59
something like precision it might be something like
11:01
precision it might be something like
11:01
precision it might be something like squared error so measuring the
11:04
squared error so measuring the
11:04
squared error so measuring the difference between the true value and in
11:07
difference between the true value and in
11:07
difference between the true value and in this case our first model we only get 30
11:09
this case our first model we only get 30
11:09
this case our first model we only get 30 percent accuracy it's not great right we
11:12
percent accuracy it's not great right we
11:12
percent accuracy it's not great right we want to improve it and so what we'll do
11:15
want to improve it and so what we'll do
11:15
want to improve it and so what we'll do is we will then put that one to the side
11:17
is we will then put that one to the side
11:17
is we will then put that one to the side then we'll iterate so we'll change all
11:19
then we'll iterate so we'll change all
11:19
then we'll iterate so we'll change all of our different values so we might
11:21
of our different values so we might
11:21
of our different values so we might choose different features a different
11:23
choose different features a different
11:23
choose different features a different algorithm set our parameters differently
11:26
algorithm set our parameters differently
11:26
algorithm set our parameters differently and then a new model comes out well we
11:29
and then a new model comes out well we
11:29
and then a new model comes out well we then have to do that over and over and
11:32
then have to do that over and over and
11:32
then have to do that over and over and over again until we finally find our 95%
11:35
over again until we finally find our 95%
11:35
over again until we finally find our 95% are really good accurate well
11:39
are really good accurate well
11:39
are really good accurate well thought-out machine learning model how
11:41
thought-out machine learning model how
11:42
thought-out machine learning model how all of that takes so much time and this
11:45
all of that takes so much time and this
11:45
all of that takes so much time and this is where automated machine learning came
11:48
is where automated machine learning came
11:48
is where automated machine learning came in to solve this problem for data
11:50
in to solve this problem for data
11:50
in to solve this problem for data scientists and so we choose to basically
11:55
scientists and so we choose to basically
11:55
scientists and so we choose to basically in automated machine learning we pass it
11:58
in automated machine learning we pass it
11:58
in automated machine learning we pass it an input so we give it our data we
12:02
an input so we give it our data we
12:02
an input so we give it our data we define the goals we want it to do so a
12:04
define the goals we want it to do so a
12:04
define the goals we want it to do so a goal might be is this a classification
12:08
goal might be is this a classification
12:08
goal might be is this a classification problem is it is something true or false
12:12
problem is it is something true or false
12:12
problem is it is something true or false is it category X Y or Z or we might do a
12:18
is it category X Y or Z or we might do a
12:18
is it category X Y or Z or we might do a regression problem so that is how much
12:20
regression problem so that is how much
12:20
regression problem so that is how much or how many so in the case of the car
12:23
or how many so in the case of the car
12:23
or how many so in the case of the car this is a regression problem how much is
12:26
this is a regression problem how much is
12:26
this is a regression problem how much is this car worth in in a in a currency
12:29
this car worth in in a in a currency
12:29
this car worth in in a in a currency amount and then we might apply some
12:31
amount and then we might apply some
12:31
amount and then we might apply some constraints some things like we can only
12:35
constraints some things like we can only
12:35
constraints some things like we can only use certain pieces of data or we can
12:38
use certain pieces of data or we can
12:38
use certain pieces of data or we can only choose certain types of algorithms
12:41
only choose certain types of algorithms
12:41
only choose certain types of algorithms that might be for policy reasons and
12:44
that might be for policy reasons and
12:44
that might be for policy reasons and that might be for training amount of
12:48
that might be for training amount of
12:48
that might be for training amount of compute reasons so for example a neural
12:50
compute reasons so for example a neural
12:50
compute reasons so for example a neural network which are very very popular are
12:53
network which are very very popular are
12:53
network which are very very popular are very powerful but they do take a long
12:55
very powerful but they do take a long
12:55
very powerful but they do take a long time to Train and if that becomes an
12:57
time to Train and if that becomes an
12:57
time to Train and if that becomes an issue sometimes you might have to apply
12:59
issue sometimes you might have to apply
12:59
issue sometimes you might have to apply that constraint but basically what we do
13:02
that constraint but basically what we do
13:02
that constraint but basically what we do then is we give it this setup and it
13:05
then is we give it this setup and it
13:05
then is we give it this setup and it will go away the automated machine
13:07
will go away the automated machine
13:07
will go away the automated machine learning service will go away it will
13:10
learning service will go away it will
13:10
learning service will go away it will test all of these different set
13:12
test all of these different set
13:12
test all of these different set oops just like we would about to do
13:15
oops just like we would about to do
13:15
oops just like we would about to do physically ourselves prior to this and
13:19
physically ourselves prior to this and
13:19
physically ourselves prior to this and then what it will do is it'll actually
13:21
then what it will do is it'll actually
13:21
then what it will do is it'll actually give us a list of all the different
13:23
give us a list of all the different
13:23
give us a list of all the different models that it was able to create as
13:25
models that it was able to create as
13:25
models that it was able to create as well as producing that final optimized
13:29
well as producing that final optimized
13:29
well as producing that final optimized model and so really like this has been a
13:35
model and so really like this has been a
13:35
model and so really like this has been a really interesting progression in
13:36
really interesting progression in
13:36
really interesting progression in machine learning for data scientists
13:38
machine learning for data scientists
13:38
machine learning for data scientists however this has also allowed a lot more
13:42
however this has also allowed a lot more
13:42
however this has also allowed a lot more people to get involved in machine
13:44
people to get involved in machine
13:44
people to get involved in machine learning as well sometimes the barrier
13:47
learning as well sometimes the barrier
13:47
learning as well sometimes the barrier to entry is the need to apply the code
13:53
to entry is the need to apply the code
13:53
to entry is the need to apply the code and the open the packages and the
13:56
and the open the packages and the
13:56
and the open the packages and the frameworks and stuff like that and you
13:58
frameworks and stuff like that and you
13:58
frameworks and stuff like that and you know that takes a long time to learn and
14:00
know that takes a long time to learn and
14:00
know that takes a long time to learn and so to actually be able to just pass it
14:04
so to actually be able to just pass it
14:04
so to actually be able to just pass it data and design your problem but not
14:07
data and design your problem but not
14:07
data and design your problem but not have to physically execute the problem
14:10
have to physically execute the problem
14:10
have to physically execute the problem it will do it for you
14:11
it will do it for you
14:11
it will do it for you this is actually meant that people who
14:14
this is actually meant that people who
14:14
this is actually meant that people who aren't experts in the machine learning
14:15
aren't experts in the machine learning
14:15
aren't experts in the machine learning space but might work with say like a
14:18
space but might work with say like a
14:18
space but might work with say like a data analyst or some you know if you
14:21
data analyst or some you know if you
14:21
data analyst or some you know if you have a kind of a data selling mindset
14:23
have a kind of a data selling mindset
14:23
have a kind of a data selling mindset and you're keen to learn about things
14:25
and you're keen to learn about things
14:25
and you're keen to learn about things like evaluation metrics and asking
14:28
like evaluation metrics and asking
14:28
like evaluation metrics and asking yourself like why is this approach to
14:31
yourself like why is this approach to
14:31
yourself like why is this approach to machine learning the right approach so
14:32
machine learning the right approach so
14:33
machine learning the right approach so am i answering my question or asking my
14:36
am i answering my question or asking my
14:36
am i answering my question or asking my question the right way and so I do think
14:38
question the right way and so I do think
14:38
question the right way and so I do think this is a really nice happy medium in
14:40
this is a really nice happy medium in
14:40
this is a really nice happy medium in some senses okay and so first things
14:45
some senses okay and so first things
14:45
some senses okay and so first things first one that's automated machine
14:47
first one that's automated machine
14:47
first one that's automated machine though so mission automated machine
14:49
though so mission automated machine
14:49
though so mission automated machine learning is the process of efficiently
14:52
learning is the process of efficiently
14:52
learning is the process of efficiently trying lots and lots of different
14:54
trying lots and lots of different
14:54
trying lots and lots of different algorithms and setups to try and solve
14:57
algorithms and setups to try and solve
14:57
algorithms and setups to try and solve your machine learning problem the
14:58
your machine learning problem the
14:58
your machine learning problem the automated piece means that it
15:01
automated piece means that it
15:01
automated piece means that it automatically runs through these and
15:03
automatically runs through these and
15:03
automatically runs through these and provides you both the optimized model as
15:05
provides you both the optimized model as
15:06
provides you both the optimized model as well as all the other models that it ran
15:08
well as all the other models that it ran
15:08
well as all the other models that it ran and from there you can go ahead and then
15:11
and from there you can go ahead and then
15:11
and from there you can go ahead and then deploy your optimized model this is also
15:14
deploy your optimized model this is also
15:14
deploy your optimized model this is also massively reduced the amount of time it
15:17
massively reduced the amount of time it
15:17
massively reduced the amount of time it takes to get models to production
15:19
takes to get models to production
15:19
takes to get models to production especially when it comes to more what we
15:23
especially when it comes to more what we
15:23
especially when it comes to more what we now call more traditional
15:25
now call more traditional
15:25
now call more traditional trista machine learning that our basic
15:27
trista machine learning that our basic
15:27
trista machine learning that our basic classification regression
15:29
classification regression
15:29
classification regression timeseriesforecasting modeling it's
15:31
timeseriesforecasting modeling it's
15:31
timeseriesforecasting modeling it's absolutely perfect to this and so what I
15:35
absolutely perfect to this and so what I
15:35
absolutely perfect to this and so what I want to do with with kind of the rest of
15:36
want to do with with kind of the rest of
15:36
want to do with with kind of the rest of the session is show you this in action
15:39
the session is show you this in action
15:39
the session is show you this in action with two different scenarios so the
15:43
with two different scenarios so the
15:43
with two different scenarios so the first scenario we're going to look at
15:44
first scenario we're going to look at
15:44
first scenario we're going to look at one one problem and then the second
15:47
one one problem and then the second
15:47
one one problem and then the second scenario we'll take a look look a quick
15:48
scenario we'll take a look look a quick
15:48
scenario we'll take a look look a quick look at time series forecasting and so
15:52
look at time series forecasting and so
15:52
look at time series forecasting and so let me introduce you to scenario number
15:55
let me introduce you to scenario number
15:55
let me introduce you to scenario number one
15:58
okay so scenario number one is from one
16:02
okay so scenario number one is from one
16:02
okay so scenario number one is from one of our favorite fictitious companies is
16:05
of our favorite fictitious companies is
16:05
of our favorite fictitious companies is called tailwind traders and Terran
16:08
called tailwind traders and Terran
16:08
called tailwind traders and Terran traders are a retail a fictitious retail
16:12
traders are a retail a fictitious retail
16:12
traders are a retail a fictitious retail company that do do-it-yourself type home
16:16
company that do do-it-yourself type home
16:16
company that do do-it-yourself type home DIY so think of things for your garden
16:20
DIY so think of things for your garden
16:20
DIY so think of things for your garden plumbing in electrical tooling stuff
16:23
plumbing in electrical tooling stuff
16:23
plumbing in electrical tooling stuff like that so there are there an online
16:25
like that so there are there an online
16:25
like that so there are there an online website as well as an in-person store
16:27
website as well as an in-person store
16:27
website as well as an in-person store and what they've had is they're the tail
16:31
and what they've had is they're the tail
16:31
and what they've had is they're the tail and traders development team have
16:33
and traders development team have
16:33
and traders development team have created this fantastic website and also
16:36
created this fantastic website and also
16:36
created this fantastic website and also a native application but as with any
16:41
a native application but as with any
16:41
a native application but as with any technology once people start to use this
16:45
technology once people start to use this
16:45
technology once people start to use this so this might be people our customers as
16:48
so this might be people our customers as
16:48
so this might be people our customers as well as partners and often people will
16:51
well as partners and often people will
16:51
well as partners and often people will find maybe issues with an application or
16:54
find maybe issues with an application or
16:54
find maybe issues with an application or a website and when they find issues they
16:57
a website and when they find issues they
16:57
a website and when they find issues they often send in support tickets or
17:00
often send in support tickets or
17:00
often send in support tickets or feedback tickets and so the tail and
17:03
feedback tickets and so the tail and
17:03
feedback tickets and so the tail and traders team have an amazing support
17:05
traders team have an amazing support
17:05
traders team have an amazing support ticket system so as these support
17:07
ticket system so as these support
17:07
ticket system so as these support tickets come in they have a great team
17:10
tickets come in they have a great team
17:10
tickets come in they have a great team who are there to triage all of the
17:14
who are there to triage all of the
17:14
who are there to triage all of the different tickets that are coming in and
17:17
different tickets that are coming in and
17:17
different tickets that are coming in and hopefully either solves them very
17:19
hopefully either solves them very
17:19
hopefully either solves them very quickly or send them on to someone who
17:21
quickly or send them on to someone who
17:21
quickly or send them on to someone who is going to act on them unfortunately
17:25
is going to act on them unfortunately
17:25
is going to act on them unfortunately these support tickets keep fly-in in and
17:28
these support tickets keep fly-in in and
17:28
these support tickets keep fly-in in and the support ticket team are starting to
17:30
the support ticket team are starting to
17:30
the support ticket team are starting to feel a little bit overwhelmed with the
17:32
feel a little bit overwhelmed with the
17:32
feel a little bit overwhelmed with the sheer amount so the more users that use
17:34
sheer amount so the more users that use
17:34
sheer amount so the more users that use it the more support tickets that come in
17:36
it the more support tickets that come in
17:36
it the more support tickets that come in and it is only a very small team
17:39
and it is only a very small team
17:39
and it is only a very small team now they did tell us that actually a lot
17:41
now they did tell us that actually a lot
17:41
now they did tell us that actually a lot of people go away very happy with the
17:43
of people go away very happy with the
17:43
of people go away very happy with the service and they get their problem
17:46
service and they get their problem
17:46
service and they get their problem solved very quickly and efficiently
17:47
solved very quickly and efficiently
17:47
solved very quickly and efficiently however unfortunately some people are
17:50
however unfortunately some people are
17:50
however unfortunately some people are not getting that great experience and so
17:53
not getting that great experience and so
17:53
not getting that great experience and so the main question here was from the tale
17:57
the main question here was from the tale
17:57
the main question here was from the tale and traders leadership team and they
17:59
and traders leadership team and they
17:59
and traders leadership team and they asked us as the data science team here
18:02
asked us as the data science team here
18:02
asked us as the data science team here they basically said this how can we help
18:05
they basically said this how can we help
18:05
they basically said this how can we help the support team that's how broad the
18:08
the support team that's how broad the
18:08
the support team that's how broad the question was and so one of the things
18:11
question was and so one of the things
18:11
question was and so one of the things around solving these problems is often
18:14
around solving these problems is often
18:14
around solving these problems is often the first question that is asked of a
18:16
the first question that is asked of a
18:16
the first question that is asked of a team to dive into data is not that it is
18:19
team to dive into data is not that it is
18:19
team to dive into data is not that it is not the question that you might
18:20
not the question that you might
18:20
not the question that you might automatically be able to apply automated
18:23
automatically be able to apply automated
18:23
automatically be able to apply automated machine learning to and so one of the
18:25
machine learning to and so one of the
18:25
machine learning to and so one of the things here is they're saying this is a
18:27
things here is they're saying this is a
18:27
things here is they're saying this is a super broad question we could help the
18:29
super broad question we could help the
18:29
super broad question we could help the support team by not needing data at all
18:32
support team by not needing data at all
18:32
support team by not needing data at all we could just say hey right we need more
18:34
we could just say hey right we need more
18:34
we could just say hey right we need more head count or more budget over there if
18:36
head count or more budget over there if
18:36
head count or more budget over there if we can find it or right we need to have
18:40
we can find it or right we need to have
18:40
we can find it or right we need to have a dedicated technical team that's going
18:42
a dedicated technical team that's going
18:42
a dedicated technical team that's going to work on this as well like it could be
18:43
to work on this as well like it could be
18:43
to work on this as well like it could be a personnel or a cultural thing and but
18:46
a personnel or a cultural thing and but
18:46
a personnel or a cultural thing and but when we started to dive into this data
18:48
when we started to dive into this data
18:48
when we started to dive into this data so we have a we have a tabular data set
18:52
so we have a we have a tabular data set
18:52
so we have a we have a tabular data set where each row in the dataset is a
18:55
where each row in the dataset is a
18:55
where each row in the dataset is a support ticket that's been submitted
18:57
support ticket that's been submitted
18:57
support ticket that's been submitted from the past and one of the things that
19:00
from the past and one of the things that
19:00
from the past and one of the things that we found was actually the duration of
19:04
we found was actually the duration of
19:04
we found was actually the duration of these support tickets it is very very
19:06
these support tickets it is very very
19:06
these support tickets it is very very interesting they solve a lot very
19:08
interesting they solve a lot very
19:08
interesting they solve a lot very quickly so between one and two days and
19:13
quickly so between one and two days and
19:13
quickly so between one and two days and then there's a bit of a dip in the
19:14
then there's a bit of a dip in the
19:14
then there's a bit of a dip in the middle so that kind of four to five days
19:16
middle so that kind of four to five days
19:16
middle so that kind of four to five days there's not a lot of support tickets
19:18
there's not a lot of support tickets
19:18
there's not a lot of support tickets sitting in there and then unfortunately
19:20
sitting in there and then unfortunately
19:20
sitting in there and then unfortunately towards the end the eight to nine day
19:22
towards the end the eight to nine day
19:22
towards the end the eight to nine day years of the graph the support ticket
19:24
years of the graph the support ticket
19:25
years of the graph the support ticket starts to rise again so there seems to
19:27
starts to rise again so there seems to
19:27
starts to rise again so there seems to be some very very quick to solve
19:28
be some very very quick to solve
19:28
be some very very quick to solve problems and then some much longer
19:31
problems and then some much longer
19:31
problems and then some much longer problems that need to be treated quite
19:33
problems that need to be treated quite
19:33
problems that need to be treated quite quickly and so one of the things we
19:36
quickly and so one of the things we
19:36
quickly and so one of the things we wanted to test was actually if we were
19:38
wanted to test was actually if we were
19:38
wanted to test was actually if we were able to predict the duration of a
19:40
able to predict the duration of a
19:40
able to predict the duration of a support ticket as it landed in the
19:43
support ticket as it landed in the
19:43
support ticket as it landed in the support ticket system could we then be
19:46
support ticket system could we then be
19:46
support ticket system could we then be much more flexible about our Stefon so
19:50
much more flexible about our Stefon so
19:50
much more flexible about our Stefon so when there's more support to get than
19:52
when there's more support to get than
19:52
when there's more support to get than or work on can we make sure that that we
19:55
or work on can we make sure that that we
19:55
or work on can we make sure that that we have kind of more stuff ready to help
19:57
have kind of more stuff ready to help
19:57
have kind of more stuff ready to help out with that so that people can quickly
20:00
out with that so that people can quickly
20:00
out with that so that people can quickly triage the the easier to solve ones and
20:04
triage the the easier to solve ones and
20:04
triage the the easier to solve ones and also spend more time working on those
20:06
also spend more time working on those
20:07
also spend more time working on those longer tickets those more complex ons
20:09
longer tickets those more complex ons
20:09
longer tickets those more complex ons and so what we're going to do is we're
20:11
and so what we're going to do is we're
20:11
and so what we're going to do is we're going to try and and we're predict the
20:16
going to try and and we're predict the
20:16
going to try and and we're predict the duration the amount of time it will take
20:18
duration the amount of time it will take
20:18
duration the amount of time it will take to solve a support ticket so let me go
20:23
to solve a support ticket so let me go
20:23
to solve a support ticket so let me go ahead and show you how we can do that so
20:27
ahead and show you how we can do that so
20:27
ahead and show you how we can do that so I'm going to I'm glad you didn't take
20:34
I'm going to I'm glad you didn't take
20:34
I'm going to I'm glad you didn't take the company name is contoso book or
20:37
the company name is contoso book or
20:37
the company name is contoso book or something else is something you always
20:41
something else is something you always
20:41
something else is something you always say quite at Microsoft contoso book and
20:44
say quite at Microsoft contoso book and
20:44
say quite at Microsoft contoso book and also menu yeah and the Italian traders
20:52
also menu yeah and the Italian traders
20:52
also menu yeah and the Italian traders is something that we've been using and
20:54
is something that we've been using and
20:54
is something that we've been using and ignite on our tour in the last couple of
20:58
ignite on our tour in the last couple of
20:58
ignite on our tour in the last couple of years and it's quite good because Tarot
21:00
years and it's quite good because Tarot
21:00
years and it's quite good because Tarot and traders are a fairly large business
21:03
and traders are a fairly large business
21:03
and traders are a fairly large business and so we can apply most technology
21:05
and so we can apply most technology
21:05
and so we can apply most technology problems to their business and in the
21:08
problems to their business and in the
21:08
problems to their business and in the data space they're actually very very
21:10
data space they're actually very very
21:10
data space they're actually very very good so it's a it's an exciting time but
21:14
good so it's a it's an exciting time but
21:14
good so it's a it's an exciting time but yes so so we are obviously as a team
21:18
yes so so we are obviously as a team
21:18
yes so so we are obviously as a team here going to be um working to try and
21:21
here going to be um working to try and
21:21
here going to be um working to try and solve that problem so first things first
21:23
solve that problem so first things first
21:23
solve that problem so first things first and I just wanted to quickly show you
21:26
and I just wanted to quickly show you
21:26
and I just wanted to quickly show you inside the azure portal if you're if
21:29
inside the azure portal if you're if
21:29
inside the azure portal if you're if you're familiar with Azure if you go to
21:31
you're familiar with Azure if you go to
21:31
you're familiar with Azure if you go to the top menu and you click create a
21:33
the top menu and you click create a
21:33
the top menu and you click create a resource and oh and you type in machine
21:45
resource and oh and you type in machine
21:45
resource and oh and you type in machine learning it will then come up with this
21:48
learning it will then come up with this
21:48
learning it will then come up with this this little soft blue canonical glass
21:51
this little soft blue canonical glass
21:51
this little soft blue canonical glass dress here and what that will mean is
21:53
dress here and what that will mean is
21:53
dress here and what that will mean is that is the the Azure machine learning
21:56
that is the the Azure machine learning
21:56
that is the the Azure machine learning service you can use a a free type tier
22:00
service you can use a a free type tier
22:00
service you can use a a free type tier so a basics here we call it and
22:05
so a basics here we call it and
22:05
so a basics here we call it and that one you can just use the SDKs and
22:08
that one you can just use the SDKs and
22:08
that one you can just use the SDKs and the notebook experiences you just pay
22:10
the notebook experiences you just pay
22:10
the notebook experiences you just pay for the compute because when using
22:12
for the compute because when using
22:12
for the compute because when using automated machine learning and it's a
22:14
automated machine learning and it's a
22:14
automated machine learning and it's a service a bit like the designer we will
22:17
service a bit like the designer we will
22:17
service a bit like the designer we will need the enterprise version of our
22:19
need the enterprise version of our
22:19
need the enterprise version of our machine learning service and if you need
22:21
machine learning service and if you need
22:21
machine learning service and if you need to find the price in do go ahead and go
22:24
to find the price in do go ahead and go
22:24
to find the price in do go ahead and go to the Asha pricing calculator and that
22:26
to the Asha pricing calculator and that
22:26
to the Asha pricing calculator and that will help you understand exactly what
22:28
will help you understand exactly what
22:28
will help you understand exactly what that means in your monthly bill but I do
22:31
that means in your monthly bill but I do
22:31
that means in your monthly bill but I do have one that I created earlier so we'll
22:34
have one that I created earlier so we'll
22:34
have one that I created earlier so we'll use that one and it's called ignite
22:36
use that one and it's called ignite
22:36
use that one and it's called ignite that's tool because it's something I've
22:37
that's tool because it's something I've
22:37
that's tool because it's something I've been using for quite a while and when
22:41
been using for quite a while and when
22:41
been using for quite a while and when you create the surface and a lot of what
22:44
you create the surface and a lot of what
22:44
you create the surface and a lot of what sits in the actual azure portal is all
22:47
sits in the actual azure portal is all
22:47
sits in the actual azure portal is all around the management of the service so
22:49
around the management of the service so
22:49
around the management of the service so thinking about who has access to your
22:51
thinking about who has access to your
22:52
thinking about who has access to your machine learning service how does that
22:54
machine learning service how does that
22:54
machine learning service how does that service have access to data sets being
22:57
service have access to data sets being
22:57
service have access to data sets being able to track and monitor the usage of
23:00
able to track and monitor the usage of
23:00
able to track and monitor the usage of this kind of stuff for Policy reasons
23:02
this kind of stuff for Policy reasons
23:02
this kind of stuff for Policy reasons stuff like that so a lot of really
23:05
stuff like that so a lot of really
23:05
stuff like that so a lot of really probably the more the IT pro type side
23:08
probably the more the IT pro type side
23:08
probably the more the IT pro type side of the house is going to spend more time
23:10
of the house is going to spend more time
23:10
of the house is going to spend more time in this part of the portal and what
23:15
in this part of the portal and what
23:15
in this part of the portal and what you'll notice is it tells you all the
23:16
you'll notice is it tells you all the
23:16
you'll notice is it tells you all the different things that it's associated
23:18
different things that it's associated
23:18
different things that it's associated with that it creates it does things like
23:20
with that it creates it does things like
23:20
with that it creates it does things like app insights you have a key vault you
23:22
app insights you have a key vault you
23:22
app insights you have a key vault you have a storage account associated with
23:24
have a storage account associated with
23:24
have a storage account associated with it and if you click on what you'll
23:26
it and if you click on what you'll
23:26
it and if you click on what you'll notice is the big house of arrow at the
23:29
notice is the big house of arrow at the
23:29
notice is the big house of arrow at the top is something of the Azure machine
23:30
top is something of the Azure machine
23:30
top is something of the Azure machine learning studio and this is probably
23:33
learning studio and this is probably
23:33
learning studio and this is probably where I spend most of my time and so
23:35
where I spend most of my time and so
23:35
where I spend most of my time and so when you click Launch now it takes you
23:38
when you click Launch now it takes you
23:38
when you click Launch now it takes you to this studio experience and it pops
23:40
to this studio experience and it pops
23:40
to this studio experience and it pops out into a separate browser window this
23:44
out into a separate browser window this
23:44
out into a separate browser window this is the place where I tend to go and
23:46
is the place where I tend to go and
23:46
is the place where I tend to go and physically do the machine learning type
23:48
physically do the machine learning type
23:48
physically do the machine learning type work and so I see this space more for
23:51
work and so I see this space more for
23:51
work and so I see this space more for the developer type of role as well as
23:54
the developer type of role as well as
23:54
the developer type of role as well as the data scientists or the data analysts
23:56
the data scientists or the data analysts
23:56
the data scientists or the data analysts so just to let you know kind of the
23:58
so just to let you know kind of the
23:58
so just to let you know kind of the difference between the two you'll spend
24:00
difference between the two you'll spend
24:00
difference between the two you'll spend most of your time actually in the Azure
24:02
most of your time actually in the Azure
24:02
most of your time actually in the Azure machine learning studio and so we're in
24:06
machine learning studio and so we're in
24:06
machine learning studio and so we're in the portal here we can see that I've
24:07
the portal here we can see that I've
24:07
the portal here we can see that I've been running some stuff prior to this
24:09
been running some stuff prior to this
24:09
been running some stuff prior to this and if we scroll down we can see I've
24:12
and if we scroll down we can see I've
24:12
and if we scroll down we can see I've got things like compute that I can
24:14
got things like compute that I can
24:14
got things like compute that I can actually attach to my experiments and so
24:17
actually attach to my experiments and so
24:17
actually attach to my experiments and so this whole as
24:18
this whole as
24:18
this whole as machine learning service is all really
24:20
machine learning service is all really
24:20
machine learning service is all really about the end-to-end and the full
24:22
about the end-to-end and the full
24:22
about the end-to-end and the full management of creating machine learning
24:24
management of creating machine learning
24:24
management of creating machine learning models so if you want to find out more
24:27
models so if you want to find out more
24:27
models so if you want to find out more about that definitely do go to something
24:29
about that definitely do go to something
24:29
about that definitely do go to something like Microsoft learn and search for
24:31
like Microsoft learn and search for
24:31
like Microsoft learn and search for Azure machine learning that'll take you
24:33
Azure machine learning that'll take you
24:33
Azure machine learning that'll take you through the real basics of you know why
24:36
through the real basics of you know why
24:36
through the real basics of you know why would I use our machine learning in this
24:38
would I use our machine learning in this
24:38
would I use our machine learning in this talk we're going to focus on just just a
24:40
talk we're going to focus on just just a
24:40
talk we're going to focus on just just a few elements of the portal so yeah I
24:42
few elements of the portal so yeah I
24:42
few elements of the portal so yeah I definitely do go and find out more and I
24:46
definitely do go and find out more and I
24:46
definitely do go and find out more and I actually see there's a there's a
24:49
actually see there's a there's a
24:49
actually see there's a there's a question that we're going to answer very
24:51
question that we're going to answer very
24:51
question that we're going to answer very shortly in the live comment so I'll be
24:53
shortly in the live comment so I'll be
24:53
shortly in the live comment so I'll be back with that question very quickly and
24:56
back with that question very quickly and
24:56
back with that question very quickly and but let me just move my laptop very
24:58
but let me just move my laptop very
24:58
but let me just move my laptop very slightly there you can see by the screen
25:00
slightly there you can see by the screen
25:00
slightly there you can see by the screen there we go so what we're going to do is
25:05
there we go so what we're going to do is
25:05
there we go so what we're going to do is we under the tab on the left so we can
25:08
we under the tab on the left so we can
25:08
we under the tab on the left so we can pop this out there we go well you can
25:12
pop this out there we go well you can
25:12
pop this out there we go well you can see we have lots of different categories
25:14
see we have lots of different categories
25:14
see we have lots of different categories so here is where we can create different
25:16
so here is where we can create different
25:16
so here is where we can create different things here they're all our assets that
25:20
things here they're all our assets that
25:20
things here they're all our assets that we actually create and this is all of
25:22
we actually create and this is all of
25:22
we actually create and this is all of our kind of management section and so if
25:25
our kind of management section and so if
25:25
our kind of management section and so if we go into assets and we take a quick
25:26
we go into assets and we take a quick
25:27
we go into assets and we take a quick look at our data sets we can see I've
25:34
look at our data sets we can see I've
25:34
look at our data sets we can see I've uploaded quite a few different data sets
25:36
uploaded quite a few different data sets
25:36
uploaded quite a few different data sets okay or you can connect to different
25:39
okay or you can connect to different
25:39
okay or you can connect to different cloud storage as well so you can
25:41
cloud storage as well so you can
25:41
cloud storage as well so you can register a datastore so that you can
25:44
register a datastore so that you can
25:44
register a datastore so that you can have a lots and lots of different and
25:48
have a lots and lots of different and
25:48
have a lots and lots of different and data sets within a single store and you
25:50
data sets within a single store and you
25:50
data sets within a single store and you don't have to move things about and have
25:52
don't have to move things about and have
25:52
don't have to move things about and have them in all different places one of the
25:54
them in all different places one of the
25:54
them in all different places one of the things you'll notice is you can do
25:55
things you'll notice is you can do
25:55
things you'll notice is you can do things like versions of data sets which
25:57
things like versions of data sets which
25:57
things like versions of data sets which is so useful with machine learning
25:58
is so useful with machine learning
25:58
is so useful with machine learning because you edit and change data sets
26:01
because you edit and change data sets
26:01
because you edit and change data sets over time and if we take a quick look at
26:03
over time and if we take a quick look at
26:03
over time and if we take a quick look at the support ticket system data
26:14
just like that dataset you'll notice
26:17
just like that dataset you'll notice
26:17
just like that dataset you'll notice that it comes up with lots of details
26:19
that it comes up with lots of details
26:19
that it comes up with lots of details all about where the data actually sits
26:21
all about where the data actually sits
26:21
all about where the data actually sits but also if you click on something like
26:23
but also if you click on something like
26:23
but also if you click on something like Explorer it can give you a very very
26:25
Explorer it can give you a very very
26:25
Explorer it can give you a very very brief summary of the actual dataset and
26:28
brief summary of the actual dataset and
26:28
brief summary of the actual dataset and it's often very useful just for
26:31
it's often very useful just for
26:31
it's often very useful just for understanding okay like how we're going
26:33
understanding okay like how we're going
26:33
understanding okay like how we're going to approach this problem what does this
26:34
to approach this problem what does this
26:34
to approach this problem what does this data look like
26:35
data look like
26:35
data look like and so here you'll see this is what our
26:38
and so here you'll see this is what our
26:39
and so here you'll see this is what our support 8 looks like we've got lots of
26:43
support 8 looks like we've got lots of
26:43
support 8 looks like we've got lots of great data in here we've got things that
26:46
great data in here we've got things that
26:46
great data in here we've got things that we do and don't want to use but
26:48
we do and don't want to use but
26:48
we do and don't want to use but basically our dataset is uploaded and
26:51
basically our dataset is uploaded and
26:51
basically our dataset is uploaded and you can do that very easily under that
26:53
you can do that very easily under that
26:53
you can do that very easily under that datasets tab but let me show you
26:56
datasets tab but let me show you
26:56
datasets tab but let me show you automated machine learning so and if we
27:00
automated machine learning so and if we
27:00
automated machine learning so and if we open our little menu here I think it
27:16
open our little menu here I think it
27:16
open our little menu here I think it might be May it does happen I don't date
27:24
might be May it does happen I don't date
27:24
might be May it does happen I don't date that I'm just I was flying through it
27:26
that I'm just I was flying through it
27:26
that I'm just I was flying through it earlier here we go right so then we can
27:28
earlier here we go right so then we can
27:28
earlier here we go right so then we can see your automated machine learning is
27:30
see your automated machine learning is
27:30
see your automated machine learning is here so if we click on the automated
27:31
here so if we click on the automated
27:31
here so if we click on the automated machine learning and it takes us to this
27:35
machine learning and it takes us to this
27:36
machine learning and it takes us to this UI page so as I mentioned you can
27:38
UI page so as I mentioned you can
27:38
UI page so as I mentioned you can actually do this via the sdk wasy but
27:41
actually do this via the sdk wasy but
27:41
actually do this via the sdk wasy but you could use code and if you feel more
27:43
you could use code and if you feel more
27:43
you could use code and if you feel more comfortable doing it that way but we're
27:45
comfortable doing it that way but we're
27:45
comfortable doing it that way but we're gonna look at the UI version which
27:47
gonna look at the UI version which
27:47
gonna look at the UI version which really takes you through step by step
27:48
really takes you through step by step
27:48
really takes you through step by step how you can approach this problem we can
27:51
how you can approach this problem we can
27:51
how you can approach this problem we can see I've run quite a lot already and we
27:53
see I've run quite a lot already and we
27:53
see I've run quite a lot already and we will revisit these because as you can
27:55
will revisit these because as you can
27:55
will revisit these because as you can see they take around 30 minutes to run
27:58
see they take around 30 minutes to run
27:58
see they take around 30 minutes to run and so I'll do a here's one I made
28:00
and so I'll do a here's one I made
28:00
and so I'll do a here's one I made earlier type thing in a moment so we'll
28:03
earlier type thing in a moment so we'll
28:03
earlier type thing in a moment so we'll go to new automated machine learning run
28:08
and you'll notice that you get kind of
28:15
and you'll notice that you get kind of
28:15
and you'll notice that you get kind of the breadcrumbs of the different areas
28:17
the breadcrumbs of the different areas
28:17
the breadcrumbs of the different areas that you're actually going to do to
28:18
that you're actually going to do to
28:18
that you're actually going to do to setup this but first things first we
28:20
setup this but first things first we
28:21
setup this but first things first we need to select our dataset so we will
28:23
need to select our dataset so we will
28:23
need to select our dataset so we will select
28:24
select
28:24
select our training data just here and what
28:30
our training data just here and what
28:30
our training data just here and what will click Next so as long as the
28:34
will click Next so as long as the
28:34
will click Next so as long as the dataset is associated in the Azure
28:36
dataset is associated in the Azure
28:36
dataset is associated in the Azure machine learning portal then we'll be
28:39
machine learning portal then we'll be
28:39
machine learning portal then we'll be able to easily select it within so right
28:43
able to easily select it within so right
28:43
able to easily select it within so right here and if it's not actually already
28:47
here and if it's not actually already
28:47
here and if it's not actually already there at this point you can also upload
28:52
there at this point you can also upload
28:52
there at this point you can also upload and create that data set so it's not
28:54
and create that data set so it's not
28:54
and create that data set so it's not like you have to go ever done this have
28:56
like you have to go ever done this have
28:56
like you have to go ever done this have I done this it does all take you through
28:59
I done this it does all take you through
28:59
I done this it does all take you through that situation as well and dataset and
29:02
that situation as well and dataset and
29:02
that situation as well and dataset and I'm gonna create a new experiment here
29:06
I'm gonna create a new experiment here
29:06
I'm gonna create a new experiment here so an experiment on for some census like
29:08
so an experiment on for some census like
29:08
so an experiment on for some census like everything around one single project you
29:11
everything around one single project you
29:11
everything around one single project you would want to keep in the same folder
29:12
would want to keep in the same folder
29:12
would want to keep in the same folder and so I think of experiment name like
29:14
and so I think of experiment name like
29:14
and so I think of experiment name like that and then we'll run lots of
29:16
that and then we'll run lots of
29:16
that and then we'll run lots of different experiments inside that one
29:19
different experiments inside that one
29:19
different experiments inside that one and experiment name and so if I enter
29:22
and experiment name and so if I enter
29:22
and experiment name and so if I enter tail and traitors support tickets at v2
29:30
and only target column this is where we
29:34
and only target column this is where we
29:34
and only target column this is where we start to define our actual automated
29:37
start to define our actual automated
29:37
start to define our actual automated machine learn and run so under target
29:39
machine learn and run so under target
29:39
machine learn and run so under target column I'll drop down and the options
29:43
column I'll drop down and the options
29:43
column I'll drop down and the options and we can see that all my different
29:44
and we can see that all my different
29:44
and we can see that all my different columns and my dates that have come up
29:46
columns and my dates that have come up
29:46
columns and my dates that have come up and we said that we want to predict the
29:49
and we said that we want to predict the
29:49
and we said that we want to predict the duration of a support ticket as it lands
29:51
duration of a support ticket as it lands
29:51
duration of a support ticket as it lands in the support ticket system so I'm
29:52
in the support ticket system so I'm
29:52
in the support ticket system so I'm going to select duration here and then
29:55
going to select duration here and then
29:55
going to select duration here and then one thing that you'll notice is that I
29:57
one thing that you'll notice is that I
29:57
one thing that you'll notice is that I then asks for some form of compute so
30:00
then asks for some form of compute so
30:00
then asks for some form of compute so down here you can set up something
30:02
down here you can set up something
30:02
down here you can set up something called machine learning compute which is
30:04
called machine learning compute which is
30:04
called machine learning compute which is a scalable form of compute so rather
30:07
a scalable form of compute so rather
30:07
a scalable form of compute so rather than a virtual machine in the cloud
30:09
than a virtual machine in the cloud
30:09
than a virtual machine in the cloud where you pay for whenever it's on and
30:13
where you pay for whenever it's on and
30:13
where you pay for whenever it's on and you have to go and deallocate it or set
30:15
you have to go and deallocate it or set
30:15
you have to go and deallocate it or set up at a line for it to deallocate itself
30:18
up at a line for it to deallocate itself
30:18
up at a line for it to deallocate itself a shutdown machine learning compute is
30:21
a shutdown machine learning compute is
30:21
a shutdown machine learning compute is very very clever in the sense that when
30:24
very very clever in the sense that when
30:24
very very clever in the sense that when it's being used it will scale up the
30:26
it's being used it will scale up the
30:26
it's being used it will scale up the nodes that it needs to use and then the
30:29
nodes that it needs to use and then the
30:29
nodes that it needs to use and then the minute they become idle after say two
30:32
minute they become idle after say two
30:32
minute they become idle after say two minutes 120 seconds it will scale it
30:35
minutes 120 seconds it will scale it
30:35
minutes 120 seconds it will scale it back down again and so you literally do
30:37
back down again and so you literally do
30:37
back down again and so you literally do only
30:38
only
30:38
only pay for the compute that it takes to run
30:40
pay for the compute that it takes to run
30:40
pay for the compute that it takes to run your experiment it's a really exciting
30:42
your experiment it's a really exciting
30:42
your experiment it's a really exciting part of the machine learning platform
30:44
part of the machine learning platform
30:44
part of the machine learning platform and it also means I no longer wake up in
30:46
and it also means I no longer wake up in
30:46
and it also means I no longer wake up in the middle of the night in hot sweats
30:48
the middle of the night in hot sweats
30:48
the middle of the night in hot sweats worrying that I have like a GPU machine
30:50
worrying that I have like a GPU machine
30:50
worrying that I have like a GPU machine left running and not doing anything so
30:53
left running and not doing anything so
30:53
left running and not doing anything so this has made my life so much easier and
30:56
this has made my life so much easier and
30:56
this has made my life so much easier and so this isn't a huge dataset and so what
30:59
so this isn't a huge dataset and so what
30:59
so this isn't a huge dataset and so what we're going to do is I have a CPU
31:01
we're going to do is I have a CPU
31:01
we're going to do is I have a CPU compute cluster so think standard
31:04
compute cluster so think standard
31:04
compute cluster so think standard developer virtual machines full nodes
31:06
developer virtual machines full nodes
31:06
developer virtual machines full nodes something like that and then I also have
31:08
something like that and then I also have
31:08
something like that and then I also have a GPU compute one available to me I'm
31:11
a GPU compute one available to me I'm
31:11
a GPU compute one available to me I'm going to CPU compute I'm going to click
31:13
going to CPU compute I'm going to click
31:13
going to CPU compute I'm going to click Next
31:14
Next
31:14
Next if you needed to create compute again
31:17
if you needed to create compute again
31:17
if you needed to create compute again you can do it during the process a
31:19
you can do it during the process a
31:19
you can do it during the process a little panel will just pop out at the
31:21
little panel will just pop out at the
31:21
little panel will just pop out at the side and you'll be able to do that ok ok
31:28
side and you'll be able to do that ok ok
31:28
side and you'll be able to do that ok ok and so now we're actually defining the
31:30
and so now we're actually defining the
31:30
and so now we're actually defining the task that we want to do so we've given
31:32
task that we want to do so we've given
31:32
task that we want to do so we've given oh no don't do that
31:35
oh no don't do that
31:35
oh no don't do that sorry I think it's lagging a little bit
31:47
apologies
31:48
apologies
31:48
apologies let me quickly try that again about time
31:54
let me quickly try that again about time
31:54
let me quickly try that again about time you doing Amy I think something that
31:55
you doing Amy I think something that
31:55
you doing Amy I think something that really interests inside there's your ml
31:57
really interests inside there's your ml
31:57
really interests inside there's your ml studio is like you started with the data
32:00
studio is like you started with the data
32:00
studio is like you started with the data set and all that without even uploading
32:01
set and all that without even uploading
32:01
set and all that without even uploading even a single excel file or CSV file
32:04
even a single excel file or CSV file
32:04
even a single excel file or CSV file which turns out Azure machine learning
32:05
which turns out Azure machine learning
32:05
which turns out Azure machine learning CMU comes with but doesn't tons and tons
32:09
CMU comes with but doesn't tons and tons
32:09
CMU comes with but doesn't tons and tons of data sets starting from very basic
32:11
of data sets starting from very basic
32:11
of data sets starting from very basic this iris did have sets to even this
32:14
this iris did have sets to even this
32:14
this iris did have sets to even this card in Titanic didn't said even there's
32:16
card in Titanic didn't said even there's
32:17
card in Titanic didn't said even there's a data set called penguin which is a
32:18
a data set called penguin which is a
32:18
a data set called penguin which is a very famous right so you have to upload
32:20
very famous right so you have to upload
32:20
very famous right so you have to upload anything you just have to select the
32:23
anything you just have to select the
32:23
anything you just have to select the data set within which you want to go and
32:25
data set within which you want to go and
32:25
data set within which you want to go and get started and there you go I think
32:27
get started and there you go I think
32:27
get started and there you go I think that is something I feel is very some
32:29
that is something I feel is very some
32:29
that is something I feel is very some something to take away from last couple
32:32
something to take away from last couple
32:32
something to take away from last couple of minutes I was gonna say and I've got
32:35
of minutes I was gonna say and I've got
32:35
of minutes I was gonna say and I've got a link actually at the end that I'll
32:36
a link actually at the end that I'll
32:36
a link actually at the end that I'll share with you that is super interesting
32:38
share with you that is super interesting
32:38
share with you that is super interesting because a lot of the notebooks and
32:41
because a lot of the notebooks and
32:41
because a lot of the notebooks and examples for the whole Azure machine
32:43
examples for the whole Azure machine
32:43
examples for the whole Azure machine learning studio all come with a scenario
32:46
learning studio all come with a scenario
32:46
learning studio all come with a scenario and a dataset that's associated with it
32:48
and a dataset that's associated with it
32:48
and a dataset that's associated with it and
32:48
and
32:48
and for example the forecasting experiment
32:51
for example the forecasting experiment
32:51
for example the forecasting experiment I'll show you I actually ran through the
32:53
I'll show you I actually ran through the
32:53
I'll show you I actually ran through the forecast an example that the Azure
32:54
forecast an example that the Azure
32:54
forecast an example that the Azure machine learning team offers and then I
32:57
machine learning team offers and then I
32:57
machine learning team offers and then I applied it afterwards to my own dataset
33:00
applied it afterwards to my own dataset
33:00
applied it afterwards to my own dataset and then that just made it like really a
33:04
and then that just made it like really a
33:04
and then that just made it like really a lot easier to know understand the
33:06
lot easier to know understand the
33:06
lot easier to know understand the approach and let me just sorry quickly
33:10
approach and let me just sorry quickly
33:10
approach and let me just sorry quickly go back to those yeah something like the
33:12
go back to those yeah something like the
33:12
go back to those yeah something like the speed that you can refer and get started
33:15
speed that you can refer and get started
33:15
speed that you can refer and get started yeah yeah and also you can easily copy
33:18
yeah yeah and also you can easily copy
33:18
yeah yeah and also you can easily copy them if you are using like the notebook
33:20
them if you are using like the notebook
33:20
them if you are using like the notebook experience or something like that you
33:22
experience or something like that you
33:22
experience or something like that you can actually copy them into your
33:24
can actually copy them into your
33:24
can actually copy them into your notebook experience because it's just
33:26
notebook experience because it's just
33:26
notebook experience because it's just like a github clone and so it copies
33:28
like a github clone and so it copies
33:28
like a github clone and so it copies everything in from that get up so it's
33:31
everything in from that get up so it's
33:31
everything in from that get up so it's really cool so apologies for that clip
33:33
really cool so apologies for that clip
33:33
really cool so apologies for that clip too soon but basically what we're going
33:35
too soon but basically what we're going
33:35
too soon but basically what we're going to do is we're going to do a regression
33:36
to do is we're going to do a regression
33:36
to do is we're going to do a regression problem so I'm going to click on
33:38
problem so I'm going to click on
33:38
problem so I'm going to click on regression here and also to note there's
33:41
regression here and also to note there's
33:41
regression here and also to note there's some additional settings that you can
33:43
some additional settings that you can
33:43
some additional settings that you can set and so things like the primary
33:46
set and so things like the primary
33:46
set and so things like the primary metric so this is how do I want it what
33:49
metric so this is how do I want it what
33:49
metric so this is how do I want it what do I want it to optimize on a metric
33:51
do I want it to optimize on a metric
33:51
do I want it to optimize on a metric wise and so for a regression problem I
33:54
wise and so for a regression problem I
33:54
wise and so for a regression problem I often use and root mean squared error
33:57
often use and root mean squared error
33:57
often use and root mean squared error it's quite a good one for and basically
34:02
it's quite a good one for and basically
34:02
it's quite a good one for and basically you do the measurement between the
34:04
you do the measurement between the
34:04
you do the measurement between the actual value and your predictive value
34:07
actual value and your predictive value
34:07
actual value and your predictive value and then you draw a line between them
34:09
and then you draw a line between them
34:09
and then you draw a line between them and the line is the amount of error so
34:11
and the line is the amount of error so
34:11
and the line is the amount of error so the closer they are together the error
34:13
the closer they are together the error
34:13
the closer they are together the error is smaller the further they are apart
34:15
is smaller the further they are apart
34:15
is smaller the further they are apart the error is much bigger and then the
34:18
the error is much bigger and then the
34:18
the error is much bigger and then the squared part of it means that when we
34:21
squared part of it means that when we
34:21
squared part of it means that when we look at say like a normal distribution
34:23
look at say like a normal distribution
34:23
look at say like a normal distribution what it does is it punishes the ones
34:26
what it does is it punishes the ones
34:26
what it does is it punishes the ones that are much further away because it
34:28
that are much further away because it
34:28
that are much further away because it squares them and so then when it's only
34:31
squares them and so then when it's only
34:31
squares them and so then when it's only small amounts it'll only square it by
34:33
small amounts it'll only square it by
34:33
small amounts it'll only square it by small amount but when it's say it's
34:36
small amount but when it's say it's
34:36
small amount but when it's say it's doing really really good in the middle
34:37
doing really really good in the middle
34:37
doing really really good in the middle of the data set but actually when it's
34:40
of the data set but actually when it's
34:40
of the data set but actually when it's on the on the extremes
34:41
on the on the extremes
34:41
on the on the extremes it's just predicting things that make no
34:44
it's just predicting things that make no
34:44
it's just predicting things that make no sense at all then it will punish it
34:46
sense at all then it will punish it
34:46
sense at all then it will punish it because it will have a much larger error
34:48
because it will have a much larger error
34:48
because it will have a much larger error and so it's a really it's a nice measure
34:51
and so it's a really it's a nice measure
34:51
and so it's a really it's a nice measure for regression but I mean all of them
34:53
for regression but I mean all of them
34:53
for regression but I mean all of them are and
34:54
are and
34:54
are and this is an interesting piece we've gone
34:57
this is an interesting piece we've gone
34:57
this is an interesting piece we've gone huge at Microsoft on responsible AI and
35:00
huge at Microsoft on responsible AI and
35:00
huge at Microsoft on responsible AI and part of responsibility is
35:02
part of responsibility is
35:02
part of responsibility is interpretability and understanding what
35:05
interpretability and understanding what
35:05
interpretability and understanding what is affecting your machine learning
35:07
is affecting your machine learning
35:07
is affecting your machine learning models and the outcome and something has
35:10
models and the outcome and something has
35:10
models and the outcome and something has been built in is this explain best model
35:12
been built in is this explain best model
35:12
been built in is this explain best model and so what it will do is it will
35:13
and so what it will do is it will
35:13
and so what it will do is it will analyze the model it's created for you
35:16
analyze the model it's created for you
35:16
analyze the model it's created for you and then you can dig into it and look at
35:19
and then you can dig into it and look at
35:19
and then you can dig into it and look at specific data points and actually start
35:22
specific data points and actually start
35:22
specific data points and actually start to understand say what columns in your
35:24
to understand say what columns in your
35:24
to understand say what columns in your dataset are heavily affects in the model
35:26
dataset are heavily affects in the model
35:26
dataset are heavily affects in the model and stuff like that
35:27
and stuff like that
35:27
and stuff like that and so you can analyze all of that great
35:28
and so you can analyze all of that great
35:28
and so you can analyze all of that great stuff and you can I mentioned you can
35:31
stuff and you can I mentioned you can
35:31
stuff and you can I mentioned you can actually block algorithms to say we
35:33
actually block algorithms to say we
35:33
actually block algorithms to say we don't have a lot of training time or a
35:35
don't have a lot of training time or a
35:35
don't have a lot of training time or a lot of compute budget for example I
35:37
lot of compute budget for example I
35:37
lot of compute budget for example I could look like neural networks or
35:39
could look like neural networks or
35:39
could look like neural networks or really big algorithms but we're not
35:42
really big algorithms but we're not
35:42
really big algorithms but we're not actually gonna do that you can do you
35:44
actually gonna do that you can do you
35:44
actually gonna do that you can do you can set the train in time so I'm going
35:46
can set the train in time so I'm going
35:46
can set the train in time so I'm going to say after an hour I think we'll have
35:49
to say after an hour I think we'll have
35:49
to say after an hour I think we'll have enough to work on and so you can set
35:51
enough to work on and so you can set
35:51
enough to work on and so you can set basically how much do I want this to
35:55
basically how much do I want this to
35:55
basically how much do I want this to really go into all of the different
35:58
really go into all of the different
35:58
really go into all of the different options that are available because if
36:00
options that are available because if
36:00
options that are available because if you've got a really big data set this
36:02
you've got a really big data set this
36:02
you've got a really big data set this could take a long time right that's the
36:04
could take a long time right that's the
36:04
could take a long time right that's the that's the payoff in some senses you're
36:08
that's the payoff in some senses you're
36:08
that's the payoff in some senses you're yes you're giving it a lot of control
36:10
yes you're giving it a lot of control
36:10
yes you're giving it a lot of control and it's doing a lot of great stuff but
36:11
and it's doing a lot of great stuff but
36:11
and it's doing a lot of great stuff but it will take time to run through all
36:13
it will take time to run through all
36:13
it will take time to run through all these experiments you can also look at
36:16
these experiments you can also look at
36:16
these experiments you can also look at things like validation and concurrency
36:18
things like validation and concurrency
36:18
things like validation and concurrency but we'll keep them as they are and then
36:21
but we'll keep them as they are and then
36:21
but we'll keep them as they are and then you can also take a look at things like
36:23
you can also take a look at things like
36:23
you can also take a look at things like the feature settings and in here you can
36:27
the feature settings and in here you can
36:27
the feature settings and in here you can actually then start to even take things
36:29
actually then start to even take things
36:29
actually then start to even take things out of your dataset that you don't want
36:32
out of your dataset that you don't want
36:32
out of your dataset that you don't want in there for this so we've tidied up our
36:35
in there for this so we've tidied up our
36:35
in there for this so we've tidied up our dataset here and actually this is
36:37
dataset here and actually this is
36:37
dataset here and actually this is everything that we want to use but say
36:39
everything that we want to use but say
36:39
everything that we want to use but say you had like so for the forecasting one
36:42
you had like so for the forecasting one
36:42
you had like so for the forecasting one oh there's a was a good example
36:43
oh there's a was a good example
36:43
oh there's a was a good example yesterday as I was working through it I
36:45
yesterday as I was working through it I
36:45
yesterday as I was working through it I was like I really don't need that that
36:49
was like I really don't need that that
36:49
was like I really don't need that that specific column in there but I really
36:51
specific column in there but I really
36:51
specific column in there but I really right now I'm just running with my
36:53
right now I'm just running with my
36:53
right now I'm just running with my testing and so I'll just remove it from
36:56
testing and so I'll just remove it from
36:56
testing and so I'll just remove it from in here and then later I looked at yeah
36:58
in here and then later I looked at yeah
36:58
in here and then later I looked at yeah update the actual dataset version and
37:00
update the actual dataset version and
37:00
update the actual dataset version and but we'll keep ours as it is and so all
37:03
but we'll keep ours as it is and so all
37:03
but we'll keep ours as it is and so all of those settings we've basically told
37:05
of those settings we've basically told
37:05
of those settings we've basically told it what dataset we've told it our
37:07
it what dataset we've told it our
37:07
it what dataset we've told it our problem
37:08
problem
37:08
problem we want to predict the duration we've
37:10
we want to predict the duration we've
37:10
we want to predict the duration we've also told it some additional things to
37:12
also told it some additional things to
37:12
also told it some additional things to optimize on and so then we'll click
37:15
optimize on and so then we'll click
37:15
optimize on and so then we'll click finish
37:16
finish
37:16
finish and it will go ahead and it'll take all
37:17
and it will go ahead and it'll take all
37:17
and it will go ahead and it'll take all of the information we've given it
37:19
of the information we've given it
37:19
of the information we've given it including our data and it will start to
37:21
including our data and it will start to
37:21
including our data and it will start to run over different models for us with
37:24
run over different models for us with
37:24
run over different models for us with different sets of parameters and
37:25
different sets of parameters and
37:25
different sets of parameters and different optimizers one of the
37:28
different optimizers one of the
37:28
different optimizers one of the questions in the chat from veena was
37:32
questions in the chat from veena was
37:32
questions in the chat from veena was does it automatically do pre-processing
37:35
does it automatically do pre-processing
37:35
does it automatically do pre-processing of input columns yes it does and so what
37:40
of input columns yes it does and so what
37:40
of input columns yes it does and so what I will show you is some of the checks
37:43
I will show you is some of the checks
37:43
I will show you is some of the checks and balances that it does for you so it
37:45
and balances that it does for you so it
37:45
and balances that it does for you so it checks things like and have you it's
37:51
checks things like and have you it's
37:51
checks things like and have you it's it's actually under the states of
37:53
it's actually under the states of
37:53
it's actually under the states of guardrails column here so once it's
37:55
guardrails column here so once it's
37:55
guardrails column here so once it's running it will start to fill out all of
37:57
running it will start to fill out all of
37:57
running it will start to fill out all of these different columns but it will look
37:59
these different columns but it will look
37:59
these different columns but it will look for things like and is your dataset have
38:02
for things like and is your dataset have
38:02
for things like and is your dataset have high cardinality so is there some data
38:05
high cardinality so is there some data
38:05
high cardinality so is there some data points that are really far away from
38:07
points that are really far away from
38:07
points that are really far away from other ones or does it have a lot of
38:10
other ones or does it have a lot of
38:10
other ones or does it have a lot of columns in it and they're quite
38:13
columns in it and they're quite
38:13
columns in it and they're quite different and it will also look for
38:15
different and it will also look for
38:15
different and it will also look for things like missing values and it can do
38:18
things like missing values and it can do
38:18
things like missing values and it can do some of its own pre-processing as well
38:20
some of its own pre-processing as well
38:20
some of its own pre-processing as well and actually when we look at the
38:23
and actually when we look at the
38:23
and actually when we look at the forecast in one it does a lot of things
38:26
forecast in one it does a lot of things
38:26
forecast in one it does a lot of things for you in that in that arena but let me
38:30
for you in that in that arena but let me
38:30
for you in that in that arena but let me show you on one that we made earlier
38:32
show you on one that we made earlier
38:32
show you on one that we made earlier what it checks on our regression problem
38:35
what it checks on our regression problem
38:35
what it checks on our regression problem and so we'll go to automated machine
38:37
and so we'll go to automated machine
38:37
and so we'll go to automated machine learning these breadcrumbs at the top
38:39
learning these breadcrumbs at the top
38:39
learning these breadcrumbs at the top are so useful instead of going back to
38:42
are so useful instead of going back to
38:42
are so useful instead of going back to like always going back to this side I
38:44
like always going back to this side I
38:44
like always going back to this side I always like go back and forth between
38:45
always like go back and forth between
38:45
always like go back and forth between the breadcrumbs because it otherwise it
38:47
the breadcrumbs because it otherwise it
38:47
the breadcrumbs because it otherwise it jumps so far that you feel like you're
38:49
jumps so far that you feel like you're
38:49
jumps so far that you feel like you're making so many clicks but and but that's
38:52
making so many clicks but and but that's
38:52
making so many clicks but and but that's a great question via you know thank you
38:54
a great question via you know thank you
38:54
a great question via you know thank you for asking that one and if you want to
38:56
for asking that one and if you want to
38:56
for asking that one and if you want to find out all the details of the
38:57
find out all the details of the
38:57
find out all the details of the pre-processing doors definitely go to
38:59
pre-processing doors definitely go to
38:59
pre-processing doors definitely go to the docs pages and inside there go to
39:06
the docs pages and inside there go to
39:06
the docs pages and inside there go to automated machines it will there's a
39:16
automated machines it will there's a
39:16
automated machines it will there's a great section on this so let me show you
39:19
great section on this so let me show you
39:19
great section on this so let me show you the docs because then that everything
39:20
the docs because then that everything
39:20
the docs because then that everything that I'll mention is sitting
39:21
that I'll mention is sitting
39:21
that I'll mention is sitting in there this one and they have a whole
39:30
in there this one and they have a whole
39:30
in there this one and they have a whole Doc's page that talks through each pro
39:33
Doc's page that talks through each pro
39:33
Doc's page that talks through each pro press pre-processing point that it does
39:35
press pre-processing point that it does
39:35
press pre-processing point that it does you can also set it and if you're using
39:38
you can also set it and if you're using
39:38
you can also set it and if you're using the SDK to not do certain pre-processing
39:40
the SDK to not do certain pre-processing
39:40
the SDK to not do certain pre-processing when you're in the UI
39:42
when you're in the UI
39:42
when you're in the UI you haven't got quite as much controls
39:44
you haven't got quite as much controls
39:44
you haven't got quite as much controls when you're using the SDK so do just
39:46
when you're using the SDK so do just
39:46
when you're using the SDK so do just keep that in mind and we'll come back to
39:51
keep that in mind and we'll come back to
39:51
keep that in mind and we'll come back to that and see if it loads and take it
39:54
that and see if it loads and take it
39:54
that and see if it loads and take it let's take a look a disservice whose
39:59
let's take a look a disservice whose
39:59
let's take a look a disservice whose combination of multiple linear
40:00
combination of multiple linear
40:00
combination of multiple linear regression models and polynomial
40:02
regression models and polynomial
40:02
regression models and polynomial regression models yes so it will choose
40:06
regression models yes so it will choose
40:06
regression models yes so it will choose a whole variety of models again there's
40:09
a whole variety of models again there's
40:09
a whole variety of models again there's a documentation page that lists every
40:11
a documentation page that lists every
40:11
a documentation page that lists every model that it tries to look at some of
40:15
model that it tries to look at some of
40:15
model that it tries to look at some of them just won't work for your problem
40:17
them just won't work for your problem
40:17
them just won't work for your problem and so it's almost a bit like of a meta
40:20
and so it's almost a bit like of a meta
40:20
and so it's almost a bit like of a meta approach because the service has
40:23
approach because the service has
40:23
approach because the service has actually learned what tends to work well
40:25
actually learned what tends to work well
40:25
actually learned what tends to work well on data and will try those algorithms
40:28
on data and will try those algorithms
40:28
on data and will try those algorithms first and that often cuts out its search
40:31
first and that often cuts out its search
40:31
first and that often cuts out its search tree on certain things and so when we
40:34
tree on certain things and so when we
40:34
tree on certain things and so when we review one of our examples when the
40:36
review one of our examples when the
40:36
review one of our examples when the interest in pieces is actually seeing
40:38
interest in pieces is actually seeing
40:38
interest in pieces is actually seeing the the types of algorithms it does but
40:44
the the types of algorithms it does but
40:44
the the types of algorithms it does but then it also changes things like the
40:46
then it also changes things like the
40:46
then it also changes things like the scaling and optimization that it runs so
40:49
scaling and optimization that it runs so
40:49
scaling and optimization that it runs so the type the approach it does to that
40:51
the type the approach it does to that
40:51
the type the approach it does to that and so you'll see like multiple of the
40:53
and so you'll see like multiple of the
40:53
and so you'll see like multiple of the same algorithm with different standard
40:56
same algorithm with different standard
40:56
same algorithm with different standard scalars and stuff like that and so yeah
40:59
scalars and stuff like that and so yeah
40:59
scalars and stuff like that and so yeah once it loads we'll take a quick look at
41:01
once it loads we'll take a quick look at
41:01
once it loads we'll take a quick look at that but that's a great question and say
41:03
that but that's a great question and say
41:03
that but that's a great question and say yes it tries a whole variety of
41:05
yes it tries a whole variety of
41:05
yes it tries a whole variety of different models and again as the
41:08
different models and again as the
41:08
different models and again as the service is progressing and maturing when
41:11
service is progressing and maturing when
41:11
service is progressing and maturing when we look at things like forecasts in in a
41:13
we look at things like forecasts in in a
41:13
we look at things like forecasts in in a moment it's using some real cutting-edge
41:15
moment it's using some real cutting-edge
41:15
moment it's using some real cutting-edge stuff like all of the ARIMA models and
41:17
stuff like all of the ARIMA models and
41:17
stuff like all of the ARIMA models and stuff like that and so definitely I
41:20
stuff like that and so definitely I
41:20
stuff like that and so definitely I think as well actually there's there's
41:23
think as well actually there's there's
41:23
think as well actually there's there's another side to this coin so and vina
41:27
another side to this coin so and vina
41:27
another side to this coin so and vina you obviously know potentially no or
41:29
you obviously know potentially no or
41:29
you obviously know potentially no or quite a bit about machine learning
41:30
quite a bit about machine learning
41:30
quite a bit about machine learning already and that's amazing but a lot of
41:33
already and that's amazing but a lot of
41:33
already and that's amazing but a lot of people might be like more like
41:35
people might be like more like
41:35
people might be like more like know what some of these words mean like
41:37
know what some of these words mean like
41:37
know what some of these words mean like don't worry it like the docs of really
41:41
don't worry it like the docs of really
41:41
don't worry it like the docs of really really good for explainer a lot of this
41:44
really good for explainer a lot of this
41:44
really good for explainer a lot of this in very very sort of clear terms that
41:49
in very very sort of clear terms that
41:49
in very very sort of clear terms that anyone can understand but I do also
41:52
anyone can understand but I do also
41:52
anyone can understand but I do also encourage that when you see some of
41:54
encourage that when you see some of
41:54
encourage that when you see some of these models and if you don't really
41:56
these models and if you don't really
41:56
these models and if you don't really understand what they are at their
41:58
understand what they are at their
41:58
understand what they are at their highest level and I recommend going and
42:01
highest level and I recommend going and
42:01
highest level and I recommend going and doing a bit of a search and kind of just
42:05
doing a bit of a search and kind of just
42:05
doing a bit of a search and kind of just spending a little bit of time to
42:06
spending a little bit of time to
42:06
spending a little bit of time to understand exactly what type of approach
42:08
understand exactly what type of approach
42:08
understand exactly what type of approach and why that might be applied best to
42:11
and why that might be applied best to
42:11
and why that might be applied best to your model and because it would be
42:13
your model and because it would be
42:13
your model and because it would be interesting how useful that is moving
42:16
interesting how useful that is moving
42:16
interesting how useful that is moving forwards as well and so yeah they're
42:19
forwards as well and so yeah they're
42:19
forwards as well and so yeah they're great questions which do deep neural
42:22
great questions which do deep neural
42:22
great questions which do deep neural network library or SDK can we use Naja
42:24
network library or SDK can we use Naja
42:24
network library or SDK can we use Naja so you can use any deep learning library
42:29
so you can use any deep learning library
42:29
so you can use any deep learning library any open source framework if you want to
42:31
any open source framework if you want to
42:31
any open source framework if you want to use PI thoughts you can use it if you
42:33
use PI thoughts you can use it if you
42:33
use PI thoughts you can use it if you want to use tensor flow you can use it
42:34
want to use tensor flow you can use it
42:34
want to use tensor flow you can use it if you only as Kerris you can use it if
42:36
if you only as Kerris you can use it if
42:36
if you only as Kerris you can use it if you want to use a specific setup of Bert
42:38
you want to use a specific setup of Bert
42:38
you want to use a specific setup of Bert that is available on github you can use
42:40
that is available on github you can use
42:40
that is available on github you can use it and the main thing here is to
42:42
it and the main thing here is to
42:42
it and the main thing here is to understand that we're looking at this
42:43
understand that we're looking at this
42:43
understand that we're looking at this from the UI but the SDK is incredibly
42:46
from the UI but the SDK is incredibly
42:46
from the UI but the SDK is incredibly powerful and what our machine learning
42:48
powerful and what our machine learning
42:48
powerful and what our machine learning is very good at is actually doing what I
42:50
is very good at is actually doing what I
42:50
is very good at is actually doing what I call the scaffolding I think of it as
42:53
call the scaffolding I think of it as
42:53
call the scaffolding I think of it as all of the stuff that sits around the
42:55
all of the stuff that sits around the
42:55
all of the stuff that sits around the project so you're trained up pyo file
42:58
project so you're trained up pyo file
42:58
project so you're trained up pyo file that you might normally create with a
43:00
that you might normally create with a
43:00
that you might normally create with a with a Python with a you know tensorflow
43:02
with a Python with a you know tensorflow
43:02
with a Python with a you know tensorflow framework or something like that
43:03
framework or something like that
43:03
framework or something like that that stays as is it's how does it run
43:07
that stays as is it's how does it run
43:07
that stays as is it's how does it run how does it do the data pre-processing
43:08
how does it do the data pre-processing
43:08
how does it do the data pre-processing step how do we get it to run on powerful
43:11
step how do we get it to run on powerful
43:11
step how do we get it to run on powerful compute in the cloud and then bring all
43:13
compute in the cloud and then bring all
43:13
compute in the cloud and then bring all those logs back to the center and so
43:16
those logs back to the center and so
43:16
those logs back to the center and so there's been a huge huge focus on making
43:19
there's been a huge huge focus on making
43:19
there's been a huge huge focus on making sure that and great frameworks that
43:21
sure that and great frameworks that
43:21
sure that and great frameworks that everyone is using in the space work in
43:24
everyone is using in the space work in
43:24
everyone is using in the space work in Asia and so that's why this kind of
43:27
Asia and so that's why this kind of
43:27
Asia and so that's why this kind of different approach almost to thinking
43:29
different approach almost to thinking
43:29
different approach almost to thinking about actually how do we make this
43:30
about actually how do we make this
43:30
about actually how do we make this production ready how do we how do you
43:32
production ready how do we how do you
43:32
production ready how do we how do you take this into real-life and manage it
43:35
take this into real-life and manage it
43:35
take this into real-life and manage it and order it and work on it as a team
43:37
and order it and work on it as a team
43:37
and order it and work on it as a team and so I see that as the as the real
43:42
and so I see that as the as the real
43:42
and so I see that as the as the real exciting part in some senses of our
43:44
exciting part in some senses of our
43:44
exciting part in some senses of our machine learning and I can only
43:46
machine learning and I can only
43:46
machine learning and I can only apologize right
43:49
my intentions to cherlene quite a lot
43:52
my intentions to cherlene quite a lot
43:52
my intentions to cherlene quite a lot here I present favor over video I could
43:58
here I present favor over video I could
43:58
here I present favor over video I could like uh try again I'm but really really
44:06
like uh try again I'm but really really
44:06
like uh try again I'm but really really great questions thank you everyone for
44:07
great questions thank you everyone for
44:07
great questions thank you everyone for asking such interesting questions and
44:17
asking such interesting questions and
44:17
asking such interesting questions and you didn't miss a lot
44:19
you didn't miss a lot
44:19
you didn't miss a lot don't worry like you can join them now
44:21
don't worry like you can join them now
44:21
don't worry like you can join them now or do some other example so we just need
44:27
or do some other example so we just need
44:27
or do some other example so we just need to take a quick look at one that I ran
44:28
to take a quick look at one that I ran
44:28
to take a quick look at one that I ran earlier that's all I need from you into
44:30
earlier that's all I need from you into
44:30
earlier that's all I need from you into that come on yeah it's taking a while
44:33
that come on yeah it's taking a while
44:33
that come on yeah it's taking a while but and let's talk a little about what
44:35
but and let's talk a little about what
44:35
but and let's talk a little about what you talked about the automation part of
44:38
you talked about the automation part of
44:38
you talked about the automation part of the machine learning right so uh if you
44:41
the machine learning right so uh if you
44:41
the machine learning right so uh if you did have this automation part a me what
44:43
did have this automation part a me what
44:43
did have this automation part a me what someone would have done is they would
44:44
someone would have done is they would
44:44
someone would have done is they would have plotted a graph maybe maybe on
44:47
have plotted a graph maybe maybe on
44:47
have plotted a graph maybe maybe on Excel they would have looked at trends
44:48
Excel they would have looked at trends
44:48
Excel they would have looked at trends on what are the possible trends and they
44:51
on what are the possible trends and they
44:51
on what are the possible trends and they would have gone to maybe another some
44:53
would have gone to maybe another some
44:53
would have gone to maybe another some ideas like spider or Python whatever OBS
44:55
ideas like spider or Python whatever OBS
44:55
ideas like spider or Python whatever OBS code they would I figured out some what
44:57
code they would I figured out some what
44:57
code they would I figured out some what are some different p-values and check
44:59
are some different p-values and check
44:59
are some different p-values and check what are the different columns that are
45:01
what are the different columns that are
45:01
what are the different columns that are not behaving properly in a machine
45:03
not behaving properly in a machine
45:03
not behaving properly in a machine learning studio so that I think for data
45:05
learning studio so that I think for data
45:05
learning studio so that I think for data scientists are data pre-processing is
45:07
scientists are data pre-processing is
45:07
scientists are data pre-processing is something that really takes a lot of
45:08
something that really takes a lot of
45:08
something that really takes a lot of time and one has to come back every
45:10
time and one has to come back every
45:11
time and one has to come back every single time when the trainer model and
45:13
single time when the trainer model and
45:13
single time when the trainer model and look at the accuracy so I think this is
45:15
look at the accuracy so I think this is
45:15
look at the accuracy so I think this is where our machine learning automation
45:17
where our machine learning automation
45:17
where our machine learning automation has it really perceived the time of a
45:20
has it really perceived the time of a
45:20
has it really perceived the time of a data scientist or any person who is
45:21
data scientist or any person who is
45:21
data scientist or any person who is working on a project is that they would
45:23
working on a project is that they would
45:23
working on a project is that they would run all the best possible algorithms are
45:27
run all the best possible algorithms are
45:27
run all the best possible algorithms are taking with the columns and they would
45:29
taking with the columns and they would
45:29
taking with the columns and they would give you the accuracy having said that
45:31
give you the accuracy having said that
45:31
give you the accuracy having said that one of the major concern is indeed the
45:33
one of the major concern is indeed the
45:33
one of the major concern is indeed the cost because computation would cost a
45:36
cost because computation would cost a
45:36
cost because computation would cost a lot but you said it would it would only
45:38
lot but you said it would it would only
45:38
lot but you said it would it would only cost you by the time it is running it
45:40
cost you by the time it is running it
45:40
cost you by the time it is running it would be people would be surely a
45:42
would be people would be surely a
45:42
would be people would be surely a concern about the costing because when
45:44
concern about the costing because when
45:44
concern about the costing because when you run so many algorithms right and
45:46
you run so many algorithms right and
45:46
you run so many algorithms right and that there's a lot of computation that
45:47
that there's a lot of computation that
45:47
that there's a lot of computation that we need and best thing as you said you
45:51
we need and best thing as you said you
45:51
we need and best thing as you said you can run any of the open source libraries
45:54
can run any of the open source libraries
45:54
can run any of the open source libraries it may be chaos tensorflow
45:56
it may be chaos tensorflow
45:56
it may be chaos tensorflow I believe it may be opening I dream to
45:58
I believe it may be opening I dream to
45:58
I believe it may be opening I dream to it you can all run that on I
46:02
it you can all run that on I
46:02
it you can all run that on I your machine learning that's great also
46:04
your machine learning that's great also
46:04
your machine learning that's great also I believe a Python comes really handy
46:07
I believe a Python comes really handy
46:07
I believe a Python comes really handy with various code
46:08
with various code
46:08
with various code Microsoft has been working really great
46:10
Microsoft has been working really great
46:11
Microsoft has been working really great with extensions that they have brought
46:12
with extensions that they have brought
46:12
with extensions that they have brought you can do the entire chip ITAT notebook
46:16
you can do the entire chip ITAT notebook
46:16
you can do the entire chip ITAT notebook inside vs code that is something that
46:18
inside vs code that is something that
46:18
inside vs code that is something that already loved so the things that are
46:22
already loved so the things that are
46:22
already loved so the things that are really amazing
46:23
really amazing
46:23
really amazing also we have Jupiters also in a local
46:25
also we have Jupiters also in a local
46:26
also we have Jupiters also in a local machine the best part of doing that
46:28
machine the best part of doing that
46:28
machine the best part of doing that Jupiter local Jupiter on Azure is like
46:30
Jupiter local Jupiter on Azure is like
46:30
Jupiter local Jupiter on Azure is like it's there it always there it doesn't
46:33
it's there it always there it doesn't
46:33
it's there it always there it doesn't matter if you come back after four years
46:35
matter if you come back after four years
46:35
matter if you come back after four years your notebooks are still there you can
46:37
your notebooks are still there you can
46:37
your notebooks are still there you can just click and run Mexican which Python
46:40
just click and run Mexican which Python
46:40
just click and run Mexican which Python value which by conversion you want to
46:42
value which by conversion you want to
46:42
value which by conversion you want to run and it's still lower this those are
46:44
run and it's still lower this those are
46:44
run and it's still lower this those are the things that I think Microsoft has
46:45
the things that I think Microsoft has
46:45
the things that I think Microsoft has really worked nicely it has been there
46:47
really worked nicely it has been there
46:47
really worked nicely it has been there for a really long time and people yet
46:50
for a really long time and people yet
46:50
for a really long time and people yet need to you know explore that part a one
46:52
need to you know explore that part a one
46:52
need to you know explore that part a one question Amy that I would like to ask
46:54
question Amy that I would like to ask
46:54
question Amy that I would like to ask you is that all that you're showing
46:56
you is that all that you're showing
46:56
you is that all that you're showing today is from a paid subscription or you
46:58
today is from a paid subscription or you
46:58
today is from a paid subscription or you have Visual Studio as a subscription
47:00
have Visual Studio as a subscription
47:00
have Visual Studio as a subscription right what about the guys who do not
47:02
right what about the guys who do not
47:02
right what about the guys who do not have the subscription how should they
47:05
have the subscription how should they
47:05
have the subscription how should they move into it because all the things that
47:07
move into it because all the things that
47:07
move into it because all the things that you have shown it's pretty fancy and
47:09
you have shown it's pretty fancy and
47:09
you have shown it's pretty fancy and anyone would like to go ahead and get
47:10
anyone would like to go ahead and get
47:10
anyone would like to go ahead and get started and yeah so number one and you
47:17
started and yeah so number one and you
47:17
started and yeah so number one and you will need an azure subscription to start
47:20
will need an azure subscription to start
47:20
will need an azure subscription to start with this so even if it's the free trial
47:22
with this so even if it's the free trial
47:22
with this so even if it's the free trial like go to an - your Microsoft comm
47:25
like go to an - your Microsoft comm
47:25
like go to an - your Microsoft comm slash free trial it's such free sorry
47:27
slash free trial it's such free sorry
47:27
slash free trial it's such free sorry and it will give you it's around two
47:30
and it will give you it's around two
47:30
and it will give you it's around two hundred dollars I think for a month
47:32
hundred dollars I think for a month
47:32
hundred dollars I think for a month and that's a lot that will really get
47:35
and that's a lot that will really get
47:35
and that's a lot that will really get you so yeah like you won't need that I
47:37
you so yeah like you won't need that I
47:37
you so yeah like you won't need that I promise you like not if my monthly bill
47:39
promise you like not if my monthly bill
47:39
promise you like not if my monthly bill was $200 I think my manager will be like
47:42
was $200 I think my manager will be like
47:42
was $200 I think my manager will be like knocking I'll be honestly like we have
47:46
knocking I'll be honestly like we have
47:46
knocking I'll be honestly like we have to be very very crude about what we use
47:48
to be very very crude about what we use
47:48
to be very very crude about what we use in our in our subscriptions just like
47:50
in our in our subscriptions just like
47:50
in our in our subscriptions just like customers do and one of the interesting
47:53
customers do and one of the interesting
47:53
customers do and one of the interesting pieces there is kind of thinking about
47:55
pieces there is kind of thinking about
47:55
pieces there is kind of thinking about that and the the pricing that sits
47:59
that and the the pricing that sits
47:59
that and the the pricing that sits behind it and so when you then come into
48:02
behind it and so when you then come into
48:02
behind it and so when you then come into the Machine and shoot if you're just
48:03
the Machine and shoot if you're just
48:03
the Machine and shoot if you're just using things like notebooks the only
48:05
using things like notebooks the only
48:05
using things like notebooks the only thing you'll pay for is the compute
48:07
thing you'll pay for is the compute
48:07
thing you'll pay for is the compute whereas if you what do want to use some
48:09
whereas if you what do want to use some
48:09
whereas if you what do want to use some of these extra services that we've built
48:13
of these extra services that we've built
48:13
of these extra services that we've built on top and they do make your life a lot
48:14
on top and they do make your life a lot
48:14
on top and they do make your life a lot easier and
48:16
easier and
48:16
easier and they are at that extra cost of the
48:18
they are at that extra cost of the
48:18
they are at that extra cost of the enterprise subscription again it is just
48:21
enterprise subscription again it is just
48:21
enterprise subscription again it is just a creating point so when you're creating
48:23
a creating point so when you're creating
48:23
a creating point so when you're creating the machine learning service in Asia you
48:25
the machine learning service in Asia you
48:25
the machine learning service in Asia you either select basic or enterprise and
48:27
either select basic or enterprise and
48:27
either select basic or enterprise and I'll be honest like I don't see it that
48:31
I'll be honest like I don't see it that
48:31
I'll be honest like I don't see it that much in my in my bill like that is not
48:36
much in my in my bill like that is not
48:36
much in my in my bill like that is not what's coming the money they're the
48:38
what's coming the money they're the
48:38
what's coming the money they're the things that cost them the money are
48:39
things that cost them the money are
48:39
things that cost them the money are things like I should container instances
48:41
things like I should container instances
48:41
things like I should container instances and as your cuba net is to deploy these
48:43
and as your cuba net is to deploy these
48:43
and as your cuba net is to deploy these models and then any virtual machines
48:45
models and then any virtual machines
48:45
models and then any virtual machines that i'm run and that's the main thing I
48:47
that i'm run and that's the main thing I
48:47
that i'm run and that's the main thing I see in my bill it's not the actual kind
48:49
see in my bill it's not the actual kind
48:49
see in my bill it's not the actual kind of payment for the for the service in
48:51
of payment for the for the service in
48:51
of payment for the for the service in some senses so do keep that in mind and
48:54
some senses so do keep that in mind and
48:54
some senses so do keep that in mind and let me let me treasure you here the one
48:57
let me let me treasure you here the one
48:57
let me let me treasure you here the one of the ones I ran I can only apologize
49:01
of the ones I ran I can only apologize
49:01
of the ones I ran I can only apologize for my internet and so sorry about that
49:03
for my internet and so sorry about that
49:03
for my internet and so sorry about that it does happen life shows it does happen
49:08
it does happen life shows it does happen
49:08
it does happen life shows it does happen and what demon always go comes in yes
49:13
and what demon always go comes in yes
49:13
and what demon always go comes in yes Lee I know I know I've such a good stuff
49:15
Lee I know I know I've such a good stuff
49:15
Lee I know I know I've such a good stuff to show you
49:16
to show you
49:16
to show you so if we don't get through it all his
49:18
so if we don't get through it all his
49:18
so if we don't get through it all his promise for you on Monday I will record
49:21
promise for you on Monday I will record
49:21
promise for you on Monday I will record all these demos and I will send them to
49:23
all these demos and I will send them to
49:23
all these demos and I will send them to you Steven and maybe you could share
49:26
you Steven and maybe you could share
49:26
you Steven and maybe you could share them with your audience as well however
49:28
them with your audience as well however
49:28
them with your audience as well however you'd like to get information out to
49:29
you'd like to get information out to
49:29
you'd like to get information out to them I said yeah I can only apologize
49:32
them I said yeah I can only apologize
49:32
them I said yeah I can only apologize sorry I need to get on the phone to my
49:34
sorry I need to get on the phone to my
49:34
sorry I need to get on the phone to my internet provider I think but here we go
49:37
internet provider I think but here we go
49:37
internet provider I think but here we go so say that we've completed and so I
49:39
so say that we've completed and so I
49:39
so say that we've completed and so I have the support ticket regression
49:41
have the support ticket regression
49:41
have the support ticket regression here's one I ran earlier and we can see
49:44
here's one I ran earlier and we can see
49:44
here's one I ran earlier and we can see that it run for 32 minutes and I ran it
49:46
that it run for 32 minutes and I ran it
49:46
that it run for 32 minutes and I ran it a few days ago so I need to know how I
49:50
a few days ago so I need to know how I
49:50
a few days ago so I need to know how I run it like what was so right I use CPU
49:52
run it like what was so right I use CPU
49:52
run it like what was so right I use CPU compute and it was done by me this is
49:55
compute and it was done by me this is
49:55
compute and it was done by me this is the dataset that it used so say you're
49:58
the dataset that it used so say you're
49:58
the dataset that it used so say you're working with a team member
49:59
working with a team member
49:59
working with a team member it's not a problem you've got all the
50:01
it's not a problem you've got all the
50:01
it's not a problem you've got all the information you need or you would need
50:02
information you need or you would need
50:02
information you need or you would need to do is I am them potentially with like
50:04
to do is I am them potentially with like
50:04
to do is I am them potentially with like anything that they were thinking and
50:08
anything that they were thinking and
50:08
anything that they were thinking and then on the left it gives you the best
50:10
then on the left it gives you the best
50:10
then on the left it gives you the best model summary and so here we can see the
50:12
model summary and so here we can see the
50:12
model summary and so here we can see the voting ensemble so an ensemble is a very
50:15
voting ensemble so an ensemble is a very
50:15
voting ensemble so an ensemble is a very powerful algorithm I'm not surprised
50:17
powerful algorithm I'm not surprised
50:17
powerful algorithm I'm not surprised that it's done so well here it's
50:19
that it's done so well here it's
50:19
that it's done so well here it's basically a mixture of multiple
50:20
basically a mixture of multiple
50:20
basically a mixture of multiple algorithms and it chooses the path that
50:23
algorithms and it chooses the path that
50:23
algorithms and it chooses the path that it wants to do that will be most optimal
50:26
it wants to do that will be most optimal
50:26
it wants to do that will be most optimal depending on the day to the point
50:28
depending on the day to the point
50:28
depending on the day to the point and then you'll see that it's given us
50:29
and then you'll see that it's given us
50:29
and then you'll see that it's given us our root mean square error it's okay
50:32
our root mean square error it's okay
50:32
our root mean square error it's okay I'll be honest it's not great for that
50:34
I'll be honest it's not great for that
50:34
I'll be honest it's not great for that to do as that and so we would want to do
50:36
to do as that and so we would want to do
50:36
to do as that and so we would want to do more tuning or we would want to look at
50:38
more tuning or we would want to look at
50:38
more tuning or we would want to look at to see actually can our dataset predict
50:40
to see actually can our dataset predict
50:40
to see actually can our dataset predict this is this a predictor is this
50:43
this is this a predictor is this
50:43
this is this a predictor is this something that we should be putting into
50:44
something that we should be putting into
50:44
something that we should be putting into production but we can also view all
50:46
production but we can also view all
50:46
production but we can also view all other metrics so it's not just the the
50:50
other metrics so it's not just the the
50:50
other metrics so it's not just the the one we choose is the only one it it
50:52
one we choose is the only one it it
50:52
one we choose is the only one it it tracks it's just the one that it's
50:54
tracks it's just the one that it's
50:54
tracks it's just the one that it's trying to optimize and so this is super
50:57
trying to optimize and so this is super
50:58
trying to optimize and so this is super interesting to see all these different
50:59
interesting to see all these different
50:59
interesting to see all these different metrics and it tells us that this
51:03
metrics and it tells us that this
51:03
metrics and it tells us that this sampling it basically took a hundred
51:06
sampling it basically took a hundred
51:06
sampling it basically took a hundred percent of our data and therefore is
51:08
percent of our data and therefore is
51:08
percent of our data and therefore is doing cross validation so for those who
51:11
doing cross validation so for those who
51:11
doing cross validation so for those who are in the space when there's training
51:14
are in the space when there's training
51:14
are in the space when there's training and testing datasets you train on a sub
51:17
and testing datasets you train on a sub
51:17
and testing datasets you train on a sub section and then you keep some of it
51:19
section and then you keep some of it
51:19
section and then you keep some of it back so that it's never seen this before
51:21
back so that it's never seen this before
51:21
back so that it's never seen this before and then you test whether it's any good
51:23
and then you test whether it's any good
51:23
and then you test whether it's any good or not and in this case it's doing cross
51:25
or not and in this case it's doing cross
51:25
or not and in this case it's doing cross validation so it will split up your full
51:28
validation so it will split up your full
51:28
validation so it will split up your full dataset so it'll take off this bit and
51:30
dataset so it'll take off this bit and
51:30
dataset so it'll take off this bit and one run and then it will take off this
51:32
one run and then it will take off this
51:32
one run and then it will take off this bit in another run and then it'll take
51:33
bit in another run and then it'll take
51:33
bit in another run and then it'll take off this bit and then this bit and then
51:35
off this bit and then this bit and then
51:35
off this bit and then this bit and then this bit and that's called five cross
51:37
this bit and that's called five cross
51:37
this bit and that's called five cross validation and we were talking about the
51:40
validation and we were talking about the
51:40
validation and we were talking about the pre-processing if you click on dates of
51:42
pre-processing if you click on dates of
51:42
pre-processing if you click on dates of guardrails when it loads this is where
51:45
guardrails when it loads this is where
51:45
guardrails when it loads this is where it tells you about the the preprocessor
51:49
it tells you about the the preprocessor
51:49
it tells you about the the preprocessor and the decisions that it's made and
51:51
and the decisions that it's made and
51:51
and the decisions that it's made and again if you go to dock start Microsoft
51:54
again if you go to dock start Microsoft
51:54
again if you go to dock start Microsoft comm and searching the Azure machine
51:55
comm and searching the Azure machine
51:55
comm and searching the Azure machine learning documentation you will see that
51:58
learning documentation you will see that
51:58
learning documentation you will see that it actually has pages for how does it do
52:01
it actually has pages for how does it do
52:01
it actually has pages for how does it do this and then when we click on the
52:03
this and then when we click on the
52:03
this and then when we click on the models tab just here this will then show
52:08
models tab just here this will then show
52:08
models tab just here this will then show us all of the different models that it
52:11
us all of the different models that it
52:11
us all of the different models that it ran in order to get to the problem so
52:13
ran in order to get to the problem so
52:13
ran in order to get to the problem so remember the reason that this was built
52:15
remember the reason that this was built
52:15
remember the reason that this was built because I would have to write code for
52:17
because I would have to write code for
52:17
because I would have to write code for just one of these track all of the
52:20
just one of these track all of the
52:20
just one of these track all of the outputs and then change it slightly and
52:22
outputs and then change it slightly and
52:22
outputs and then change it slightly and rerun it and then change it slightly and
52:24
rerun it and then change it slightly and
52:25
rerun it and then change it slightly and rerun it and then change it slightly
52:26
rerun it and then change it slightly
52:26
rerun it and then change it slightly every eyes a lot of time like that takes
52:29
every eyes a lot of time like that takes
52:29
every eyes a lot of time like that takes so long that takes some look way more
52:31
so long that takes some look way more
52:31
so long that takes some look way more than 30 minutes of my time so this it is
52:35
than 30 minutes of my time so this it is
52:35
than 30 minutes of my time so this it is incredible what it can do and so here
52:38
incredible what it can do and so here
52:38
incredible what it can do and so here you'll see that it's ranking them by or
52:40
you'll see that it's ranking them by or
52:40
you'll see that it's ranking them by or normalized root mean squared error
52:42
normalized root mean squared error
52:42
normalized root mean squared error but I mean look at the amount that it
52:44
but I mean look at the amount that it
52:44
but I mean look at the amount that it runs it runs about in general up to like
52:48
runs it runs about in general up to like
52:49
runs it runs about in general up to like 30 each time it does it and so it's
52:51
30 each time it does it and so it's
52:51
30 each time it does it and so it's fascinating to see those different runs
52:53
fascinating to see those different runs
52:53
fascinating to see those different runs and you can also dig into like the logs
52:56
and you can also dig into like the logs
52:56
and you can also dig into like the logs and see actually how it ran so if you're
52:58
and see actually how it ran so if you're
52:58
and see actually how it ran so if you're if you normally code in the space I
53:00
if you normally code in the space I
53:00
if you normally code in the space I recommend looking at that but when we go
53:03
recommend looking at that but when we go
53:03
recommend looking at that but when we go to the details page and we choose our
53:05
to the details page and we choose our
53:05
to the details page and we choose our best algorithm one of the interesting
53:08
best algorithm one of the interesting
53:08
best algorithm one of the interesting parts of bad fits that will click on
53:10
parts of bad fits that will click on
53:10
parts of bad fits that will click on voting ensemble we can actually dive
53:12
voting ensemble we can actually dive
53:12
voting ensemble we can actually dive into that specific model and when we do
53:16
into that specific model and when we do
53:16
into that specific model and when we do it will tell us a little bit more about
53:21
it will tell us a little bit more about
53:21
it will tell us a little bit more about it we I mentioned the interpret ability
53:24
it we I mentioned the interpret ability
53:24
it we I mentioned the interpret ability this is where you would see it so you
53:26
this is where you would see it so you
53:26
this is where you would see it so you can click on the excavations tab and it
53:28
can click on the excavations tab and it
53:28
can click on the excavations tab and it will start to load up your
53:29
will start to load up your
53:29
will start to load up your interpretability your explain your best
53:32
interpretability your explain your best
53:32
interpretability your explain your best model so thinking about thinking about
53:36
model so thinking about thinking about
53:36
model so thinking about thinking about you know these the why of why you're
53:40
you know these the why of why you're
53:40
you know these the why of why you're doing this and actually what in the
53:42
doing this and actually what in the
53:42
doing this and actually what in the dataset is really pushing it but while I
53:45
dataset is really pushing it but while I
53:45
dataset is really pushing it but while I show you which you can literally click
53:47
show you which you can literally click
53:47
show you which you can literally click so say this is right you're like this is
53:49
so say this is right you're like this is
53:49
so say this is right you're like this is great this is exactly what I need and I
53:51
great this is exactly what I need and I
53:51
great this is exactly what I need and I need to get something out into a into an
53:55
need to get something out into a into an
53:55
need to get something out into a into an environment that people can actually use
53:56
environment that people can actually use
53:56
environment that people can actually use it you can click on deploy here and it
54:00
it you can click on deploy here and it
54:00
it you can click on deploy here and it will then deploy this to an azure
54:03
will then deploy this to an azure
54:03
will then deploy this to an azure container instance so we can give it a
54:05
container instance so we can give it a
54:05
container instance so we can give it a name so we can say throw in trailers
54:07
name so we can say throw in trailers
54:07
name so we can say throw in trailers support ticket rush free to just let it
54:15
support ticket rush free to just let it
54:15
support ticket rush free to just let it catch up with me just there there we go
54:17
catch up with me just there there we go
54:17
catch up with me just there there we go you can give it a description you can
54:19
you can give it a description you can
54:19
you can give it a description you can you choose the compute types that IKS is
54:22
you choose the compute types that IKS is
54:22
you choose the compute types that IKS is asha kubernetes service and a CI is asha
54:26
asha kubernetes service and a CI is asha
54:26
asha kubernetes service and a CI is asha container instances container instances
54:29
container instances container instances
54:29
container instances container instances are a single instance of a container it
54:32
are a single instance of a container it
54:32
are a single instance of a container it can scale a little bit and stuff like
54:34
can scale a little bit and stuff like
54:34
can scale a little bit and stuff like that but we recommend for production and
54:36
that but we recommend for production and
54:36
that but we recommend for production and workloads you look at using Azure
54:38
workloads you look at using Azure
54:38
workloads you look at using Azure kubernetes service because of all the
54:40
kubernetes service because of all the
54:40
kubernetes service because of all the amazing manageability that comes with it
54:42
amazing manageability that comes with it
54:42
amazing manageability that comes with it but given that we're just wearing betta
54:44
but given that we're just wearing betta
54:44
but given that we're just wearing betta here we're just testing I will use that
54:47
here we're just testing I will use that
54:47
here we're just testing I will use that and what's it not like
54:49
and what's it not like
54:49
and what's it not like this strange pol appears that's four yes
55:07
this strange pol appears that's four yes
55:07
this strange pol appears that's four yes you can enable authentication so one of
55:11
you can enable authentication so one of
55:11
you can enable authentication so one of the key things right is machine learning
55:13
the key things right is machine learning
55:13
the key things right is machine learning models one of the ways that they can be
55:15
models one of the ways that they can be
55:15
models one of the ways that they can be attacked in technology is actually by
55:18
attacked in technology is actually by
55:18
attacked in technology is actually by people sending rubbish data to them and
55:21
people sending rubbish data to them and
55:21
people sending rubbish data to them and having the access to do so and then it
55:23
having the access to do so and then it
55:23
having the access to do so and then it will just like mess it all up and then
55:25
will just like mess it all up and then
55:25
will just like mess it all up and then that data might be going into a new data
55:28
that data might be going into a new data
55:28
that data might be going into a new data store for you to retrain the model and
55:30
store for you to retrain the model and
55:30
store for you to retrain the model and you can look at omission around that and
55:32
you can look at omission around that and
55:32
you can look at omission around that and all of sudden your model is just going
55:34
all of sudden your model is just going
55:34
all of sudden your model is just going you know crazy so it's super interesting
55:37
you know crazy so it's super interesting
55:37
you know crazy so it's super interesting to see that you can add or enable
55:40
to see that you can add or enable
55:40
to see that you can add or enable authentication and you can really think
55:41
authentication and you can really think
55:41
authentication and you can really think this is an API endpoint so just like any
55:44
this is an API endpoint so just like any
55:44
this is an API endpoint so just like any API endpoint we must be able to protect
55:50
API endpoint we must be able to protect
55:50
API endpoint we must be able to protect it it's just like any other container
55:52
it it's just like any other container
55:52
it it's just like any other container right we wouldn't leave another
55:53
right we wouldn't leave another
55:53
right we wouldn't leave another container out there with those well in
55:55
container out there with those well in
55:55
container out there with those well in this case we are not going to enable
55:57
this case we are not going to enable
55:57
this case we are not going to enable authentication but you would click
55:58
authentication but you would click
55:59
authentication but you would click deploy just here and you can add and
56:01
deploy just here and you can add and
56:01
deploy just here and you can add and actually add your own custom assets as
56:03
actually add your own custom assets as
56:03
actually add your own custom assets as well so if you want to add custom
56:04
well so if you want to add custom
56:04
well so if you want to add custom scoring scripts because you need to do a
56:06
scoring scripts because you need to do a
56:06
scoring scripts because you need to do a little bit of processing at the end you
56:08
little bit of processing at the end you
56:08
little bit of processing at the end you can do that here we won't do that
56:11
can do that here we won't do that
56:11
can do that here we won't do that because I have here's one I made earlier
56:12
because I have here's one I made earlier
56:12
because I have here's one I made earlier type thing so when you have created one
56:15
type thing so when you have created one
56:15
type thing so when you have created one it takes about 15 to 20 minutes normally
56:18
it takes about 15 to 20 minutes normally
56:18
it takes about 15 to 20 minutes normally until it's fully ready it will have it
56:21
until it's fully ready it will have it
56:21
until it's fully ready it will have it we spinning it up and transitioning it
56:24
we spinning it up and transitioning it
56:24
we spinning it up and transitioning it and then and putting your model into it
56:26
and then and putting your model into it
56:26
and then and putting your model into it etc when you go into endpoints you can
56:30
etc when you go into endpoints you can
56:30
etc when you go into endpoints you can see the endpoints so the container
56:32
see the endpoints so the container
56:32
see the endpoints so the container instances here that I have and when I
56:34
instances here that I have and when I
56:34
instances here that I have and when I click in on them it then becomes what
56:37
click in on them it then becomes what
56:37
click in on them it then becomes what reminds me a little bit of like
56:38
reminds me a little bit of like
56:38
reminds me a little bit of like cognitive services so it gives you an
56:40
cognitive services so it gives you an
56:40
cognitive services so it gives you an endpoint that you need to call or score
56:42
endpoint that you need to call or score
56:42
endpoint that you need to call or score II in URI and it tells you the schema
56:47
II in URI and it tells you the schema
56:47
II in URI and it tells you the schema that you need to call it with the data
56:51
that you need to call it with the data
56:51
that you need to call it with the data schema and then you go ahead and you can
56:54
schema and then you go ahead and you can
56:54
schema and then you go ahead and you can actually use that and so just while this
56:58
actually use that and so just while this
56:58
actually use that and so just while this loads what really
57:01
loads what really
57:01
loads what really question and a huge thing right now low
57:03
question and a huge thing right now low
57:03
question and a huge thing right now low code development is mainly in the past
57:07
code development is mainly in the past
57:07
code development is mainly in the past we've done a well how do I expand so
57:10
we've done a well how do I expand so
57:10
we've done a well how do I expand so I've seen this is some trends in the
57:13
I've seen this is some trends in the
57:13
I've seen this is some trends in the time I've been in tech and I'm sure many
57:15
time I've been in tech and I'm sure many
57:15
time I've been in tech and I'm sure many people here have been in take way longer
57:17
people here have been in take way longer
57:17
people here have been in take way longer than me so I'm sure you'll see it as
57:18
than me so I'm sure you'll see it as
57:18
than me so I'm sure you'll see it as well especially in the data science of
57:21
well especially in the data science of
57:21
well especially in the data science of machine learning space things start in
57:23
machine learning space things start in
57:23
machine learning space things start in research and we're talking research code
57:26
research and we're talking research code
57:26
research and we're talking research code as well we're talking like quite meta
57:29
as well we're talking like quite meta
57:29
as well we're talking like quite meta level heavy theory-based
57:32
level heavy theory-based
57:32
level heavy theory-based not ready for production right research
57:35
not ready for production right research
57:35
not ready for production right research code is kind of different it's proving a
57:37
code is kind of different it's proving a
57:37
code is kind of different it's proving a point it's proven that something's
57:38
point it's proven that something's
57:38
point it's proven that something's possible to do and then what we do is we
57:41
possible to do and then what we do is we
57:41
possible to do and then what we do is we take research code and we bring it
57:43
take research code and we bring it
57:43
take research code and we bring it closer to production and then we
57:46
closer to production and then we
57:46
closer to production and then we actually say well you know what we want
57:48
actually say well you know what we want
57:48
actually say well you know what we want to be doing here is we want to be trying
57:52
to be doing here is we want to be trying
57:52
to be doing here is we want to be trying to get it to become real and often that
57:54
to get it to become real and often that
57:54
to get it to become real and often that becomes some kind a framework or SDK or
57:58
becomes some kind a framework or SDK or
57:58
becomes some kind a framework or SDK or maybe a package that you that you
58:01
maybe a package that you that you
58:01
maybe a package that you that you actually have and you use but you have
58:03
actually have and you use but you have
58:03
actually have and you use but you have to understand the the depths of it in
58:07
to understand the the depths of it in
58:07
to understand the the depths of it in some senses to actually use it and then
58:09
some senses to actually use it and then
58:09
some senses to actually use it and then we see higher level things like we see
58:12
we see higher level things like we see
58:12
we see higher level things like we see things that were built into visual
58:15
things that were built into visual
58:15
things that were built into visual studio code that will help you pull
58:17
studio code that will help you pull
58:17
studio code that will help you pull together samples of scripts or a
58:19
together samples of scripts or a
58:19
together samples of scripts or a framework that actually does so much for
58:21
framework that actually does so much for
58:21
framework that actually does so much for you and you just call methods and insert
58:24
you and you just call methods and insert
58:24
you and you just call methods and insert variables and it does it but actually
58:26
variables and it does it but actually
58:26
variables and it does it but actually behind those frameworks there's a lot of
58:29
behind those frameworks there's a lot of
58:29
behind those frameworks there's a lot of complex code and then there's this next
58:32
complex code and then there's this next
58:32
complex code and then there's this next step right and I almost see it as like
58:35
step right and I almost see it as like
58:35
step right and I almost see it as like aisles to pass to sass in some senses
58:38
aisles to pass to sass in some senses
58:38
aisles to pass to sass in some senses it's almost like taking it each time a
58:41
it's almost like taking it each time a
58:41
it's almost like taking it each time a little bit more responsibility away from
58:43
little bit more responsibility away from
58:43
little bit more responsibility away from you and soullow code is now super
58:46
you and soullow code is now super
58:46
you and soullow code is now super interesting because it means that
58:48
interesting because it means that
58:48
interesting because it means that actually if I don't like I'm acting I
58:52
actually if I don't like I'm acting I
58:52
actually if I don't like I'm acting I can code like no worries but if I have
58:54
can code like no worries but if I have
58:54
can code like no worries but if I have has to like why why would I
58:57
has to like why why would I
58:57
has to like why why would I I know it sounds really really bad but
58:59
I know it sounds really really bad but
58:59
I know it sounds really really bad but like why would out if I can get there
59:01
like why would out if I can get there
59:01
like why would out if I can get there quicker by just using a tool that's
59:03
quicker by just using a tool that's
59:03
quicker by just using a tool that's gonna help me do that
59:05
gonna help me do that
59:05
gonna help me do that then then why wouldn't I but at it but
59:07
then then why wouldn't I but at it but
59:07
then then why wouldn't I but at it but also when it gets too low code it opens
59:09
also when it gets too low code it opens
59:09
also when it gets too low code it opens up the area for everyone to come and
59:12
up the area for everyone to come and
59:12
up the area for everyone to come and join and be able to use it
59:14
join and be able to use it
59:14
join and be able to use it and then it means that we don't have to
59:15
and then it means that we don't have to
59:15
and then it means that we don't have to have because we don't have enough
59:17
have because we don't have enough
59:17
have because we don't have enough experts right in the world we just don't
59:19
experts right in the world we just don't
59:19
experts right in the world we just don't especially in the machine learning space
59:22
especially in the machine learning space
59:22
especially in the machine learning space - for the demands that's needed and so
59:24
- for the demands that's needed and so
59:24
- for the demands that's needed and so it's for the ability of people being
59:26
it's for the ability of people being
59:26
it's for the ability of people being able to create these amazing solutions
59:28
able to create these amazing solutions
59:28
able to create these amazing solutions that help people use it that's why I see
59:31
that help people use it that's why I see
59:31
that help people use it that's why I see low code has been so interesting but
59:34
low code has been so interesting but
59:34
low code has been so interesting but back to where we were we had an endpoint
59:36
back to where we were we had an endpoint
59:36
back to where we were we had an endpoint so an azure container incident sitting
59:39
so an azure container incident sitting
59:39
so an azure container incident sitting in the cloud I had our automated machine
59:42
in the cloud I had our automated machine
59:42
in the cloud I had our automated machine learning and mobile on it and it has
59:44
learning and mobile on it and it has
59:44
learning and mobile on it and it has this rest point here as well as a
59:46
this rest point here as well as a
59:46
this rest point here as well as a swaggered definition so if I open up
59:48
swaggered definition so if I open up
59:48
swaggered definition so if I open up swagger swagger is an open source
59:51
swagger swagger is an open source
59:51
swagger swagger is an open source standard there is all four rest api's
59:54
standard there is all four rest api's
59:54
standard there is all four rest api's and and how you call them and if we copy
59:59
and and how you call them and if we copy
59:59
and and how you call them and if we copy our rest endpoint just here and let me
1:00:04
our rest endpoint just here and let me
1:00:04
our rest endpoint just here and let me open postman sorry I should have had
1:00:06
open postman sorry I should have had
1:00:06
open postman sorry I should have had that one open earlier but basically what
1:00:33
that one open earlier but basically what
1:00:33
that one open earlier but basically what we can do is we can hopefully this is
1:00:35
we can do is we can hopefully this is
1:00:35
we can do is we can hopefully this is loaded this one won't look very nice
1:00:37
loaded this one won't look very nice
1:00:37
loaded this one won't look very nice this will just be a huge JSON string
1:00:40
this will just be a huge JSON string
1:00:40
this will just be a huge JSON string when it changes and that is telling us
1:00:44
when it changes and that is telling us
1:00:44
when it changes and that is telling us what I breach a schema is there we go so
1:00:48
what I breach a schema is there we go so
1:00:48
what I breach a schema is there we go so it's a bit of a mess and so one of the
1:00:50
it's a bit of a mess and so one of the
1:00:50
it's a bit of a mess and so one of the little neat little tricks that I tend to
1:00:52
little neat little tricks that I tend to
1:00:52
little neat little tricks that I tend to do is I actually take the whole swagger
1:00:57
do is I actually take the whole swagger
1:00:57
do is I actually take the whole swagger definition I open visual studio code
1:01:02
definition I open visual studio code
1:01:02
definition I open visual studio code paste it in there we go and something
1:01:08
paste it in there we go and something
1:01:08
paste it in there we go and something that you can do is you can do an alt
1:01:11
that you can do is you can do an alt
1:01:11
that you can do is you can do an alt shift and s and it will actually she
1:01:16
shift and s and it will actually she
1:01:16
shift and s and it will actually she says just good so that's it all
1:01:30
[Music]
1:01:33
right now I want to go away and why is
1:01:37
right now I want to go away and why is
1:01:37
right now I want to go away and why is they all shift and therefore then they
1:01:39
they all shift and therefore then they
1:01:39
they all shift and therefore then they then lay out really nicely and it's so
1:01:41
then lay out really nicely and it's so
1:01:41
then lay out really nicely and it's so much easier to read like this and if you
1:01:44
much easier to read like this and if you
1:01:44
much easier to read like this and if you scroll down in this schema definition
1:01:46
scroll down in this schema definition
1:01:46
scroll down in this schema definition it's well worth taking a read through
1:01:48
it's well worth taking a read through
1:01:48
it's well worth taking a read through your schema definition in detail but one
1:01:50
your schema definition in detail but one
1:01:50
your schema definition in detail but one of the things that you'll see just after
1:01:52
of the things that you'll see just after
1:01:52
of the things that you'll see just after this here
1:01:53
this here
1:01:53
this here is it will give you an example of the
1:01:55
is it will give you an example of the
1:01:55
is it will give you an example of the data input and so you can then select
1:01:58
data input and so you can then select
1:01:58
data input and so you can then select this pop it into your postpone request
1:02:03
this pop it into your postpone request
1:02:03
this pop it into your postpone request your REST API request and check that
1:02:07
your REST API request and check that
1:02:07
your REST API request and check that works that's what I do first I just put
1:02:08
works that's what I do first I just put
1:02:08
works that's what I do first I just put the default data in in and then
1:02:10
the default data in in and then
1:02:10
the default data in in and then understand the schema and then I can
1:02:12
understand the schema and then I can
1:02:12
understand the schema and then I can start pulling together say like a okay
1:02:14
start pulling together say like a okay
1:02:14
start pulling together say like a okay well what does that look like when I'm
1:02:15
well what does that look like when I'm
1:02:15
well what does that look like when I'm having to put like many many different
1:02:17
having to put like many many different
1:02:17
having to put like many many different things or how do I build back into my
1:02:18
things or how do I build back into my
1:02:18
things or how do I build back into my code in some senses and and we have
1:02:21
code in some senses and and we have
1:02:21
code in some senses and and we have postman here just here that's that's
1:02:26
postman here just here that's that's
1:02:26
postman here just here that's that's kind of just loading for me but one of
1:02:32
kind of just loading for me but one of
1:02:32
kind of just loading for me but one of the things I mainly just like to show
1:02:34
the things I mainly just like to show
1:02:34
the things I mainly just like to show you though is it does just become a rest
1:02:36
you though is it does just become a rest
1:02:36
you though is it does just become a rest endpoint and that's the key thing of it
1:02:40
endpoint and that's the key thing of it
1:02:40
endpoint and that's the key thing of it it's not having meaning that you need to
1:02:42
it's not having meaning that you need to
1:02:42
it's not having meaning that you need to build in a certain SDK into your code or
1:02:45
build in a certain SDK into your code or
1:02:45
build in a certain SDK into your code or anything like that if you can make a
1:02:46
anything like that if you can make a
1:02:46
anything like that if you can make a rest call you can call now your bespoke
1:02:49
rest call you can call now your bespoke
1:02:49
rest call you can call now your bespoke machine learning algorithm that has been
1:02:51
machine learning algorithm that has been
1:02:51
machine learning algorithm that has been built and run against multiple different
1:02:54
built and run against multiple different
1:02:54
built and run against multiple different sources in multiple different algorithms
1:02:56
sources in multiple different algorithms
1:02:56
sources in multiple different algorithms and parameters to actually choose you
1:02:59
and parameters to actually choose you
1:02:59
and parameters to actually choose you the best one and very very quickly just
1:03:03
the best one and very very quickly just
1:03:03
the best one and very very quickly just for the end and I absolutely will send
1:03:06
for the end and I absolutely will send
1:03:06
for the end and I absolutely will send some videos out because this has not
1:03:08
some videos out because this has not
1:03:08
some videos out because this has not been as slick apologies that I would
1:03:10
been as slick apologies that I would
1:03:10
been as slick apologies that I would have liked it I would say more to show
1:03:12
have liked it I would say more to show
1:03:12
have liked it I would say more to show you so I will definitely get those out
1:03:14
you so I will definitely get those out
1:03:14
you so I will definitely get those out to you and actually they're really
1:03:15
to you and actually they're really
1:03:15
to you and actually they're really useful them for when if you want to step
1:03:17
useful them for when if you want to step
1:03:17
useful them for when if you want to step along with me you can pause it and stuff
1:03:20
along with me you can pause it and stuff
1:03:20
along with me you can pause it and stuff and if we go to forecasting forecasting
1:03:23
and if we go to forecasting forecasting
1:03:23
and if we go to forecasting forecasting is a new piece in this space so it's
1:03:27
is a new piece in this space so it's
1:03:27
is a new piece in this space so it's something that was added more recently
1:03:29
something that was added more recently
1:03:29
something that was added more recently and there is some amazing so the focus
1:03:31
and there is some amazing so the focus
1:03:31
and there is some amazing so the focus is actually very very hard but I I think
1:03:33
is actually very very hard but I I think
1:03:33
is actually very very hard but I I think it's quite a hard problem to solve you
1:03:35
it's quite a hard problem to solve you
1:03:35
it's quite a hard problem to solve you have to think a lot about all the things
1:03:37
have to think a lot about all the things
1:03:37
have to think a lot about all the things compared to the other algorithms
1:03:38
compared to the other algorithms
1:03:38
compared to the other algorithms think of things like lag and how one
1:03:41
think of things like lag and how one
1:03:41
think of things like lag and how one variable associates to another in a time
1:03:44
variable associates to another in a time
1:03:44
variable associates to another in a time series and it also uses very very
1:03:47
series and it also uses very very
1:03:47
series and it also uses very very different algorithms for these kinds of
1:03:48
different algorithms for these kinds of
1:03:48
different algorithms for these kinds of approaches and so one of the things and
1:03:51
approaches and so one of the things and
1:03:51
approaches and so one of the things and that I found is actually automated
1:03:52
that I found is actually automated
1:03:52
that I found is actually automated machine learning has really done a great
1:03:54
machine learning has really done a great
1:03:54
machine learning has really done a great job of of teaching me a lot more about
1:03:57
job of of teaching me a lot more about
1:03:57
job of of teaching me a lot more about forecasting and forecasting is so useful
1:04:02
forecasting and forecasting is so useful
1:04:02
forecasting and forecasting is so useful in businesses right thinking forecasting
1:04:05
in businesses right thinking forecasting
1:04:05
in businesses right thinking forecasting sales forecasting demands forecasting
1:04:08
sales forecasting demands forecasting
1:04:08
sales forecasting demands forecasting need for something etc it's all very
1:04:13
need for something etc it's all very
1:04:13
need for something etc it's all very very important and we can see here
1:04:16
very important and we can see here
1:04:16
very important and we can see here that's good to run a hundred that was my
1:04:19
that's good to run a hundred that was my
1:04:19
that's good to run a hundred that was my last one I am able to set up the problem
1:04:23
last one I am able to set up the problem
1:04:23
last one I am able to set up the problem but all it does is give me a few extra
1:04:26
but all it does is give me a few extra
1:04:26
but all it does is give me a few extra options and I'll select and so yeah I'll
1:04:29
options and I'll select and so yeah I'll
1:04:29
options and I'll select and so yeah I'll record something up for you on this
1:04:30
record something up for you on this
1:04:30
record something up for you on this piece around forecasting and we can kind
1:04:34
piece around forecasting and we can kind
1:04:34
piece around forecasting and we can kind of dig a little deeper into that in
1:04:36
of dig a little deeper into that in
1:04:36
of dig a little deeper into that in there but it all looks the same and so
1:04:39
there but it all looks the same and so
1:04:39
there but it all looks the same and so you get the same setup so the
1:04:42
you get the same setup so the
1:04:42
you get the same setup so the interesting thing when I was learning
1:04:43
interesting thing when I was learning
1:04:43
interesting thing when I was learning about forecasting bit was because
1:04:45
about forecasting bit was because
1:04:45
about forecasting bit was because classification and regression are very
1:04:48
classification and regression are very
1:04:48
classification and regression are very well practiced in that space I felt very
1:04:51
well practiced in that space I felt very
1:04:51
well practiced in that space I felt very comfortable that the step into
1:04:53
comfortable that the step into
1:04:53
comfortable that the step into forecasting was a lot smaller I did have
1:04:56
forecasting was a lot smaller I did have
1:04:56
forecasting was a lot smaller I did have to do quite a bit of reading around like
1:04:58
to do quite a bit of reading around like
1:04:58
to do quite a bit of reading around like some of the the different types of
1:05:02
some of the the different types of
1:05:02
some of the the different types of approaches that are used etc and what
1:05:04
approaches that are used etc and what
1:05:04
approaches that are used etc and what some of these variables mean when it's
1:05:06
some of these variables mean when it's
1:05:06
some of these variables mean when it's asking me to set them but other than
1:05:08
asking me to set them but other than
1:05:08
asking me to set them but other than that it does a very very good job and so
1:05:11
that it does a very very good job and so
1:05:11
that it does a very very good job and so it does things like it automatically
1:05:12
it does things like it automatically
1:05:12
it does things like it automatically detects the step in your date time so in
1:05:17
detects the step in your date time so in
1:05:17
detects the step in your date time so in your time series so you have an IOT
1:05:19
your time series so you have an IOT
1:05:19
your time series so you have an IOT devices your actual series might be
1:05:22
devices your actual series might be
1:05:22
devices your actual series might be every minute whereas in a transactional
1:05:25
every minute whereas in a transactional
1:05:25
every minute whereas in a transactional database your series might be every day
1:05:27
database your series might be every day
1:05:27
database your series might be every day or something like that and so it's able
1:05:31
or something like that and so it's able
1:05:31
or something like that and so it's able to detect that automatically it brings
1:05:34
to detect that automatically it brings
1:05:34
to detect that automatically it brings in all of the forecasting the lag and
1:05:37
in all of the forecasting the lag and
1:05:37
in all of the forecasting the lag and the understanding like how it predicts
1:05:41
the understanding like how it predicts
1:05:41
the understanding like how it predicts each value depending on what you're
1:05:44
each value depending on what you're
1:05:44
each value depending on what you're trying to forecast and so I'll
1:05:45
trying to forecast and so I'll
1:05:45
trying to forecast and so I'll definitely send you some more stuff but
1:05:47
definitely send you some more stuff but
1:05:47
definitely send you some more stuff but you can absolutely do that with
1:05:49
you can absolutely do that with
1:05:49
you can absolutely do that with automated machine learning as well
1:05:52
automated machine learning as well
1:05:52
automated machine learning as well and so I guess because we I'm a little
1:05:54
and so I guess because we I'm a little
1:05:54
and so I guess because we I'm a little bit over and so I want to be very very
1:05:56
bit over and so I want to be very very
1:05:56
bit over and so I want to be very very conscious of your time and I just want
1:05:59
conscious of your time and I just want
1:05:59
conscious of your time and I just want to share with you some of our amazing
1:06:01
to share with you some of our amazing
1:06:01
to share with you some of our amazing links so just here and if you want to
1:06:06
links so just here and if you want to
1:06:06
links so just here and if you want to keep learning as I said I will send out
1:06:08
keep learning as I said I will send out
1:06:08
keep learning as I said I will send out some videos soon I'll send them to you
1:06:10
some videos soon I'll send them to you
1:06:10
some videos soon I'll send them to you and as soon as possible on Monday and
1:06:13
and as soon as possible on Monday and
1:06:13
and as soon as possible on Monday and the the best place to go is to go to the
1:06:16
the the best place to go is to go to the
1:06:16
the the best place to go is to go to the documentation I've brought it up over
1:06:18
documentation I've brought it up over
1:06:18
documentation I've brought it up over and over again there it's a KA OMS slash
1:06:22
and over again there it's a KA OMS slash
1:06:22
and over again there it's a KA OMS slash mythbuster - Auto ml so that will take
1:06:25
mythbuster - Auto ml so that will take
1:06:25
mythbuster - Auto ml so that will take you to a place where you can start
1:06:27
you to a place where you can start
1:06:27
you to a place where you can start understanding automating machine
1:06:30
understanding automating machine
1:06:30
understanding automating machine learning more I did mention
1:06:31
learning more I did mention
1:06:31
learning more I did mention understanding the metrics I believe is
1:06:34
understanding the metrics I believe is
1:06:34
understanding the metrics I believe is incredibly important so I've actually
1:06:35
incredibly important so I've actually
1:06:35
incredibly important so I've actually got your short link there specifically
1:06:37
got your short link there specifically
1:06:37
got your short link there specifically to the page that does understanding of
1:06:39
to the page that does understanding of
1:06:39
to the page that does understanding of metrics so that's a kms such auto am l -
1:06:42
metrics so that's a kms such auto am l -
1:06:42
metrics so that's a kms such auto am l - metrics and then for all of the Azure
1:06:45
metrics and then for all of the Azure
1:06:45
metrics and then for all of the Azure machine learning documentation and it's
1:06:48
machine learning documentation and it's
1:06:48
machine learning documentation and it's at that link there so that's doc stop
1:06:50
at that link there so that's doc stop
1:06:50
at that link there so that's doc stop Microsoft comm such a sure slash machine
1:06:53
Microsoft comm such a sure slash machine
1:06:53
Microsoft comm such a sure slash machine - learning that's where you'll get
1:06:55
- learning that's where you'll get
1:06:55
- learning that's where you'll get everything in that Azure machine
1:06:56
everything in that Azure machine
1:06:57
everything in that Azure machine learning portal but it also recommend
1:06:59
learning portal but it also recommend
1:06:59
learning portal but it also recommend going to Microsoft learn and because as
1:07:01
going to Microsoft learn and because as
1:07:02
going to Microsoft learn and because as the Azure machine learning platform has
1:07:03
the Azure machine learning platform has
1:07:03
the Azure machine learning platform has a few different modules in there that
1:07:05
a few different modules in there that
1:07:05
a few different modules in there that might be a great source to have a quick
1:07:08
might be a great source to have a quick
1:07:08
might be a great source to have a quick look over and I also wanted to share
1:07:11
look over and I also wanted to share
1:07:11
look over and I also wanted to share I've started to do this more because I
1:07:13
I've started to do this more because I
1:07:13
I've started to do this more because I think it's really important and some
1:07:15
think it's really important and some
1:07:15
think it's really important and some influences in the spaces of other people
1:07:17
influences in the spaces of other people
1:07:17
influences in the spaces of other people that I know are working in this space
1:07:19
that I know are working in this space
1:07:19
that I know are working in this space and so I have a few people for you here
1:07:21
and so I have a few people for you here
1:07:21
and so I have a few people for you here these are these are their twitter
1:07:23
these are these are their twitter
1:07:23
these are these are their twitter handles so Francesca is on our team she
1:07:27
handles so Francesca is on our team she
1:07:27
handles so Francesca is on our team she is an absolute expert in data science
1:07:31
is an absolute expert in data science
1:07:31
is an absolute expert in data science and she's also a very very savvy
1:07:32
and she's also a very very savvy
1:07:32
and she's also a very very savvy full-custom person she is the person
1:07:34
full-custom person she is the person
1:07:34
full-custom person she is the person I've learned a lot about for custom from
1:07:37
I've learned a lot about for custom from
1:07:37
I've learned a lot about for custom from we also have re re Bornstein he's also
1:07:40
we also have re re Bornstein he's also
1:07:40
we also have re re Bornstein he's also on our team and he's very very into
1:07:42
on our team and he's very very into
1:07:42
on our team and he's very very into things like natural language processing
1:07:44
things like natural language processing
1:07:44
things like natural language processing but it's an absolute superstar around
1:07:46
but it's an absolute superstar around
1:07:46
but it's an absolute superstar around the Azure machine learning platform and
1:07:48
the Azure machine learning platform and
1:07:48
the Azure machine learning platform and some of its deep technical and elements
1:07:51
some of its deep technical and elements
1:07:51
some of its deep technical and elements around the SDK and then also a gentleman
1:07:54
around the SDK and then also a gentleman
1:07:54
around the SDK and then also a gentleman in the community called Vlad and he did
1:07:58
in the community called Vlad and he did
1:07:58
in the community called Vlad and he did this amazing talk it's a part of the
1:08:00
this amazing talk it's a part of the
1:08:00
this amazing talk it's a part of the global AI community which I've just put
1:08:02
global AI community which I've just put
1:08:02
global AI community which I've just put at the bottom
1:08:04
at the bottom
1:08:04
at the bottom and he did an amazing talk where he used
1:08:06
and he did an amazing talk where he used
1:08:06
and he did an amazing talk where he used automated machine learning to help him
1:08:09
automated machine learning to help him
1:08:09
automated machine learning to help him with a cattle composition and so
1:08:11
with a cattle composition and so
1:08:11
with a cattle composition and so calibration
1:08:12
calibration
1:08:12
calibration yeah and it's it's such a good talk so
1:08:14
yeah and it's it's such a good talk so
1:08:14
yeah and it's it's such a good talk so go check out the lab and go check out
1:08:17
go check out the lab and go check out
1:08:17
go check out the lab and go check out the global AI community as well I know
1:08:19
the global AI community as well I know
1:08:19
the global AI community as well I know CV I know you're a big supporter of the
1:08:21
CV I know you're a big supporter of the
1:08:21
CV I know you're a big supporter of the community and in there on the YouTube
1:08:26
community and in there on the YouTube
1:08:26
community and in there on the YouTube channel for the globe layer community
1:08:27
channel for the globe layer community
1:08:27
channel for the globe layer community they actually have Lud's talk so I
1:08:29
they actually have Lud's talk so I
1:08:29
they actually have Lud's talk so I highly recommend going to take a look at
1:08:31
highly recommend going to take a look at
1:08:31
highly recommend going to take a look at that as well you forgot to add your name
1:08:34
that as well you forgot to add your name
1:08:34
that as well you forgot to add your name Amy so yeah if you want to follow me
1:08:43
Amy so yeah if you want to follow me
1:08:43
Amy so yeah if you want to follow me that would be an absolute treat I tend
1:08:44
that would be an absolute treat I tend
1:08:44
that would be an absolute treat I tend to share pretty much just a really broad
1:08:49
to share pretty much just a really broad
1:08:49
to share pretty much just a really broad spectrum I've done things in designer
1:08:51
spectrum I've done things in designer
1:08:51
spectrum I've done things in designer Enloe code as well as looking at things
1:08:54
Enloe code as well as looking at things
1:08:54
Enloe code as well as looking at things like our SDK so and yeah a bit of a
1:08:57
like our SDK so and yeah a bit of a
1:08:57
like our SDK so and yeah a bit of a broad brush approach but often just the
1:08:59
broad brush approach but often just the
1:08:59
broad brush approach but often just the ability to use the technology is a place
1:09:02
ability to use the technology is a place
1:09:02
ability to use the technology is a place that I like to kind of show people like
1:09:04
that I like to kind of show people like
1:09:04
that I like to kind of show people like that kind of element yeah so one day me
1:09:11
that kind of element yeah so one day me
1:09:11
that kind of element yeah so one day me that did cover that he said that it was
1:09:12
that did cover that he said that it was
1:09:12
that did cover that he said that it was in the paid subscription but you can
1:09:15
in the paid subscription but you can
1:09:15
in the paid subscription but you can somehow manage and go get started even
1:09:17
somehow manage and go get started even
1:09:17
somehow manage and go get started even with the you know it's great in debit
1:09:20
with the you know it's great in debit
1:09:20
with the you know it's great in debit card but they do have a Azure machine
1:09:22
card but they do have a Azure machine
1:09:22
card but they do have a Azure machine learning classic right if not all they
1:09:25
learning classic right if not all they
1:09:25
learning classic right if not all they are options available at least you can
1:09:27
are options available at least you can
1:09:27
are options available at least you can go and get with either side you get all
1:09:28
go and get with either side you get all
1:09:28
go and get with either side you get all those algorithms and you can also go and
1:09:30
those algorithms and you can also go and
1:09:30
those algorithms and you can also go and get started by create once you create
1:09:32
get started by create once you create
1:09:32
get started by create once you create your machine learning model and then you
1:09:34
your machine learning model and then you
1:09:34
your machine learning model and then you go and create those api's it does allow
1:09:37
go and create those api's it does allow
1:09:37
go and create those api's it does allow you to create aps on the fly it could
1:09:39
you to create aps on the fly it could
1:09:39
you to create aps on the fly it could add the input and the output node all on
1:09:41
add the input and the output node all on
1:09:41
add the input and the output node all on its own
1:09:42
its own
1:09:42
its own so even if you do not don't want to go
1:09:44
so even if you do not don't want to go
1:09:44
so even if you do not don't want to go and give your credentials you can still
1:09:46
and give your credentials you can still
1:09:46
and give your credentials you can still go and with a simple outlook on you can
1:09:48
go and with a simple outlook on you can
1:09:48
go and with a simple outlook on you can get started with the classic as you
1:09:50
get started with the classic as you
1:09:50
get started with the classic as you machine an in-studio well and and and I
1:09:53
machine an in-studio well and and and I
1:09:53
machine an in-studio well and and and I think Amy that was a really great
1:09:55
think Amy that was a really great
1:09:55
think Amy that was a really great session that you put along right it was
1:09:58
session that you put along right it was
1:09:58
session that you put along right it was an amazing situation we started with
1:10:00
an amazing situation we started with
1:10:00
an amazing situation we started with what is actually automation with machine
1:10:02
what is actually automation with machine
1:10:02
what is actually automation with machine learning you went on to explain the with
1:10:05
learning you went on to explain the with
1:10:05
learning you went on to explain the with examples right how it saves you time
1:10:07
examples right how it saves you time
1:10:07
examples right how it saves you time with pre-processing then of course then
1:10:10
with pre-processing then of course then
1:10:10
with pre-processing then of course then K the demo daemon it is for but I'm
1:10:16
K the demo daemon it is for but I'm
1:10:16
K the demo daemon it is for but I'm I'm so sorry I owe you a shame I have
1:10:19
I'm so sorry I owe you a shame I have
1:10:19
I'm so sorry I owe you a shame I have some amazing to find out I can apologize
1:10:22
some amazing to find out I can apologize
1:10:22
some amazing to find out I can apologize I hope it's been useful to just hit here
1:10:24
I hope it's been useful to just hit here
1:10:24
I hope it's been useful to just hit here it talked through we we do get through
1:10:27
it talked through we we do get through
1:10:27
it talked through we we do get through kind of that that oneself however as I
1:10:30
kind of that that oneself however as I
1:10:30
kind of that that oneself however as I said I'll do a demo record and I'll send
1:10:33
said I'll do a demo record and I'll send
1:10:33
said I'll do a demo record and I'll send it out and and that will be a lot
1:10:36
it out and and that will be a lot
1:10:36
it out and and that will be a lot quicker through the process it will give
1:10:38
quicker through the process it will give
1:10:38
quicker through the process it will give you that sense of ends ones but I hope
1:10:39
you that sense of ends ones but I hope
1:10:39
you that sense of ends ones but I hope this conversation has also been useful
1:10:42
this conversation has also been useful
1:10:42
this conversation has also been useful and thank you so much for your questions
1:10:44
and thank you so much for your questions
1:10:44
and thank you so much for your questions as well they've all been really really
1:10:45
as well they've all been really really
1:10:45
as well they've all been really really good that's great that's great so thank
1:10:47
good that's great that's great so thank
1:10:48
good that's great that's great so thank you everyone who has joined us for today
1:10:49
you everyone who has joined us for today
1:10:49
you everyone who has joined us for today I think we reached really good audience
1:10:51
I think we reached really good audience
1:10:51
I think we reached really good audience and thanks what went was asked questions
1:10:53
and thanks what went was asked questions
1:10:53
and thanks what went was asked questions so Amy let's just wrap it up I think I
1:10:56
so Amy let's just wrap it up I think I
1:10:56
so Amy let's just wrap it up I think I had a gray eye and the anti viewers had
1:10:59
had a gray eye and the anti viewers had
1:10:59
had a gray eye and the anti viewers had a really great time the I Steven Simon
1:11:01
a really great time the I Steven Simon
1:11:01
a really great time the I Steven Simon on behalf of anti sea shark on our
1:11:04
on behalf of anti sea shark on our
1:11:04
on behalf of anti sea shark on our community and its millions of users
1:11:06
community and its millions of users
1:11:06
community and its millions of users would like to thank you for your
1:11:07
would like to thank you for your
1:11:07
would like to thank you for your valuable time and they're going to be in
1:11:09
valuable time and they're going to be in
1:11:09
valuable time and they're going to be in stores the community we would love to
1:11:12
stores the community we would love to
1:11:12
stores the community we would love to have you back makes whenever you are
1:11:14
have you back makes whenever you are
1:11:14
have you back makes whenever you are available and looking forward thank you
1:11:18
available and looking forward thank you
1:11:18
available and looking forward thank you thank you stay happy and enjoy your
1:11:21
thank you stay happy and enjoy your
1:11:21
thank you stay happy and enjoy your weekend thank you

Automated Machine Learning with Low Code - AMA

CSharpCorner

AI for Developers: Enhancing Skills, Not Replacing - MVP Show ft. Chris Love

The Outer Worlds 2 - A Complication With The Computronic Cerebrum: Resolve the Personality Crisis

Optimizing Email Triage Using Azure AI Language Studio and Azure Logic Apps

Oliver & Company | DISNEY THIS DAY | November 18, 1988

ABANDONED BUILDINGS & FORGOTTEN SPACES: Horror Lurks in the Shadows!

Ready for a Baby? But Not for Marriage?Jhene Aiko and Big Sean as case study

Flock: Technology's Role in Law Enforcement & Safety #shorts

Neuralink Revolution: Merging Science Fiction with Reality #Singularity #AivsAi #Metaverse

Reality Stars Who Refused To Get Close To Each Other

TV Moments That Went Way Too Far

Tides of Annihilation - Official Break The Mirror EXTENDED CUT Gameplay Trailer

Geoguessr But I Have Infinite Guesses...

Risk of Rain 2: Alloyed Collective - Official Launch Trailer

I Learned the Shape of EVERY European Country

My Amazon account got hacked. I'll tell you what they did.

Up next in 10

Automated Machine Learning with Low Code - AMA

CSharpCorner