I’m Kaz Sato I’m a developer advocate
I’m Kaz Sato I’m a developer advocate for Google cloud platform and I’m also a
for Google cloud platform and I’m also a
for Google cloud platform and I’m also a tech lead for the data analytics team
tech lead for the data analytics team
tech lead for the data analytics team and I have been working at Google for
and I have been working at Google for
and I have been working at Google for over five years and for the last one and
over five years and for the last one and
over five years and for the last one and a half years I have been working as a
a half years I have been working as a
a half years I have been working as a developer advocate to having a
developer advocate to having a
developer advocate to having a presentation like this at many events so
presentation like this at many events so
presentation like this at many events so in these sessions I’d like to talk about
in these sessions I’d like to talk about
in these sessions I’d like to talk about these agendas but first I’d like to
these agendas but first I’d like to
these agendas but first I’d like to introduction introduce the concept of
introduction introduce the concept of
introduction introduce the concept of neural network and deep learning and how
neural network and deep learning and how
neural network and deep learning and how it works with very great demonstrations
it works with very great demonstrations
it works with very great demonstrations and also I’d like to introduce how
and also I’d like to introduce how
and also I’d like to introduce how Google has been deploying those neural
Google has been deploying those neural
Google has been deploying those neural network technologies to the Google crowd
network technologies to the Google crowd
network technologies to the Google crowd and then I’ll be covering the
and then I’ll be covering the
and then I’ll be covering the technologies and products that those are
technologies and products that those are
technologies and products that those are actually provided as a products from
actually provided as a products from
actually provided as a products from Google cloud platforms so what what is
Google cloud platforms so what what is
Google cloud platforms so what what is neural network so neural network is a
neural network so neural network is a
neural network so neural network is a function that can learn from training
function that can learn from training
function that can learn from training data set so if you want to have newer
data set so if you want to have newer
data set so if you want to have newer networks to do image recognition then
networks to do image recognition then
networks to do image recognition then you can put the for example cat image is
you can put the for example cat image is
you can put the for example cat image is converted into a load vector and then
converted into a load vector and then
converted into a load vector and then put that vector into the networks then
put that vector into the networks then
put that vector into the networks then eventually you’d have another output
eventually you’d have another output
eventually you’d have another output vector which represents the labels of
vector which represents the labels of
vector which represents the labels of the objects detect such as cat or human
the objects detect such as cat or human
the objects detect such as cat or human face so it’s designed to mimic the
face so it’s designed to mimic the
face so it’s designed to mimic the behavior of new ones inside human brains
behavior of new ones inside human brains
behavior of new ones inside human brains by using matrix operations so actually
by using matrix operations so actually
by using matrix operations so actually it’s really it has very basic matrix
it’s really it has very basic matrix
it’s really it has very basic matrix operations only there is no
operations only there is no
operations only there is no sophisticated or fancy mathematics going
sophisticated or fancy mathematics going
sophisticated or fancy mathematics going on everything you do with neural
on everything you do with neural
on everything you do with neural networks is you have if you matrix
networks is you have if you matrix
networks is you have if you matrix operations we have learned at high
operations we have learned at high
operations we have learned at high school for example the input vector may
school for example the input vector may
school for example the input vector may we present represent the image of cat
we present represent the image of cat
we present represent the image of cat then they will have a vector that would
then they will have a vector that would
then they will have a vector that would have the pixel data converting it to a
have the pixel data converting it to a
have the pixel data converting it to a vet
vet
vet and you would get another output vector
and you would get another output vector
and you would get another output vector as a result that represents the label of
as a result that represents the label of
as a result that represents the label of the detected images in in this case you
the detected images in in this case you
the detected images in in this case you will have the any number closer to the
will have the any number closer to the
will have the any number closer to the 1.0 that indicates the neural networks
1.0 that indicates the neural networks
1.0 that indicates the neural networks thinks the image must be a cat but you
thinks the image must be a cat but you
thinks the image must be a cat but you think’s going on inside in your networks
think’s going on inside in your networks
think’s going on inside in your networks is very simple all you are doing with
is very simple all you are doing with
is very simple all you are doing with neural network is the matrix operations
neural network is the matrix operations
neural network is the matrix operations like W X plus B equals wise where the WS
like W X plus B equals wise where the WS
like W X plus B equals wise where the WS are weights and B are biases and
are weights and B are biases and
are weights and B are biases and actually you don’t have to care about
actually you don’t have to care about
actually you don’t have to care about those WS and B’s at all you let
those WS and B’s at all you let
those WS and B’s at all you let computers find and calculate the WS and
computers find and calculate the WS and
computers find and calculate the WS and B’s so all you have to care is the yeah
B’s so all you have to care is the yeah
B’s so all you have to care is the yeah what kind of data you would want to put
what kind of data you would want to put
what kind of data you would want to put into the networks and what kind of the
into the networks and what kind of the
into the networks and what kind of the result you want to get so let’s take a
result you want to get so let’s take a
result you want to get so let’s take a look at the some interesting
look at the some interesting
look at the some interesting demonstration of neural networks for
demonstration of neural networks for
demonstration of neural networks for example if you have a problem like this
example if you have a problem like this
example if you have a problem like this the very simple classifications how you
the very simple classifications how you
the very simple classifications how you can use neural networks to do the
can use neural networks to do the
can use neural networks to do the classifications of these two different
classifications of these two different
classifications of these two different data sets I’m not sure what that those
data sets I’m not sure what that those
data sets I’m not sure what that those data point means but let’s take a lift
data point means but let’s take a lift
data point means but let’s take a lift imagine that those are the data of the
imagine that those are the data of the
imagine that those are the data of the weights and height of people so that if
weights and height of people so that if
weights and height of people so that if a person’s weight and height are both
a person’s weight and height are both
a person’s weight and height are both loud and you can think he or she must be
loud and you can think he or she must be
loud and you can think he or she must be a child or maybe he or she could be a
a child or maybe he or she could be a
a child or maybe he or she could be a adult how how can you classify them if
adult how how can you classify them if
adult how how can you classify them if you using you if you try to use neural
you using you if you try to use neural
you using you if you try to use neural networks to do this problem then you can
networks to do this problem then you can
networks to do this problem then you can just apply the same equation W X plus B
just apply the same equation W X plus B
just apply the same equation W X plus B equal Y to do you’ve to do two
equal Y to do you’ve to do two
equal Y to do you’ve to do two classifications so you are putting the
classifications so you are putting the
classifications so you are putting the weight number and height number here as
weight number and height number here as
weight number and height number here as a vector and you get an output a vector
a vector and you get an output a vector
a vector and you get an output a vector like this where if you if what single
like this where if you if what single
like this where if you if what single data points isn’t classified as an adult
data points isn’t classified as an adult
data points isn’t classified as an adult then you would have one here and if it’s
then you would have one here and if it’s
then you would have one here and if it’s a child then you would have a one here
a child then you would have a one here
a child then you would have a one here and thing is that computer tries to find
and thing is that computer tries to find
and thing is that computer tries to find the optimal combination of the
the optimal combination of the
the optimal combination of the parameters such as the weights and
parameters such as the weights and
parameters such as the weights and biases by itself and you don’t have to
biases by itself and you don’t have to
biases by itself and you don’t have to think about what kind of parameters you
think about what kind of parameters you
think about what kind of parameters you have to set to neural networks computer
have to set to neural networks computer
have to set to neural networks computer does it for you so let’s take a look at
does it for you so let’s take a look at
does it for you so let’s take a look at the actual demonstration did you see
the actual demonstration did you see
the actual demonstration did you see that so I can do that okay so instead of
that so I can do that okay so instead of
that so I can do that okay so instead of the humans letting computer instructing
the humans letting computer instructing
the humans letting computer instructing learning you know you don’t have to
learning you know you don’t have to
learning you know you don’t have to teach computers how to solve these
teach computers how to solve these
teach computers how to solve these problems but everything you have to do
problems but everything you have to do
problems but everything you have to do is provide the training data set so that
is provide the training data set so that
is provide the training data set so that computer thinks by itself to optimize
computer thinks by itself to optimize
computer thinks by itself to optimize the combination of the parameters such
the combination of the parameters such
the combination of the parameters such as weights and biases so you see that
as weights and biases so you see that
as weights and biases so you see that the computer is try to change the the
the computer is try to change the the
the computer is try to change the the weights to do the classifications at
weights to do the classifications at
weights to do the classifications at optimal success rate and now computer is
optimal success rate and now computer is
optimal success rate and now computer is using an algorithm is called gradient
using an algorithm is called gradient
using an algorithm is called gradient lee set that means it tries to increase
lee set that means it tries to increase
lee set that means it tries to increase or decrease each weights and biases to
or decrease each weights and biases to
or decrease each weights and biases to make the combinations closer through the
make the combinations closer through the
make the combinations closer through the higher accuracy or lower loss slate so
higher accuracy or lower loss slate so
higher accuracy or lower loss slate so it’s just like we’re learning things
it’s just like we’re learning things
it’s just like we’re learning things from the the parents or there may be a
from the the parents or there may be a
from the the parents or there may be a senior people in your company where you
senior people in your company where you
senior people in your company where you this junior people or these children are
this junior people or these children are
this junior people or these children are learning from many many mistakes so
learning from many many mistakes so
learning from many many mistakes so computer makes many many mistakes but in
computer makes many many mistakes but in
computer makes many many mistakes but in the initial stage but if you provide
the initial stage but if you provide
the initial stage but if you provide much much more training data set than
much much more training data set than
much much more training data set than computer using the gradient descent
computer using the gradient descent
computer using the gradient descent algorithms try to minimize the failures
algorithms try to minimize the failures
algorithms try to minimize the failures so that that’s how it works so let’s
so that that’s how it works so let’s
so that that’s how it works so let’s take a look at another interesting
take a look at another interesting
take a look at another interesting demonstrations of new or networks where
demonstrations of new or networks where
demonstrations of new or networks where you have the another training dataset
you have the another training dataset
you have the another training dataset like this I’m not sure what what does it
like this I’m not sure what what does it
like this I’m not sure what what does it mean but we think that we have a some
mean but we think that we have a some
mean but we think that we have a some data set which requires a complex
data set which requires a complex
data set which requires a complex classifications if you a program has
classifications if you a program has
classifications if you a program has maybe with this data setting you may
maybe with this data setting you may
maybe with this data setting you may want to user maybe equations for circle
want to user maybe equations for circle
want to user maybe equations for circle to classify the two datasets through
to classify the two datasets through
to classify the two datasets through datasets and arrange a disease but by
datasets and arrange a disease but by
datasets and arrange a disease but by using newer networks you can just let
using newer networks you can just let
using newer networks you can just let computer things how to solve it now you
computer things how to solve it now you
computer things how to solve it now you saw the computer was trying to create a
saw the computer was trying to create a
saw the computer was trying to create a pattern to classify those datasets by
pattern to classify those datasets by
pattern to classify those datasets by using so called neurons in the hidden
using so called neurons in the hidden
using so called neurons in the hidden layers with the first example we didn’t
layers with the first example we didn’t
layers with the first example we didn’t have any hidden layers but with this
have any hidden layers but with this
have any hidden layers but with this complex data that you would need to use
complex data that you would need to use
complex data that you would need to use the hidden layers that means between the
the hidden layers that means between the
the hidden layers that means between the input data and the output neurons you
input data and the output neurons you
input data and the output neurons you would have another layers between them
would have another layers between them
would have another layers between them there has much pool new ones and each
there has much pool new ones and each
there has much pool new ones and each new ones does very simple things these
new ones does very simple things these
new ones does very simple things these neurons only classifies whether the data
neurons only classifies whether the data
neurons only classifies whether the data points is in the in the bottom left red
points is in the in the bottom left red
points is in the in the bottom left red area or the upper right area or now
area or the upper right area or now
area or the upper right area or now these neurons only classifies whether
these neurons only classifies whether
these neurons only classifies whether the data points is in the left or right
the data points is in the left or right
the data points is in the left or right just like that but combining those
just like that but combining those
just like that but combining those outputs at the the last new one neural
outputs at the the last new one neural
outputs at the the last new one neural network then the neural networks can
network then the neural networks can
network then the neural networks can compose much more complex pattern like
compose much more complex pattern like
compose much more complex pattern like this and if you have more and more new
this and if you have more and more new
this and if you have more and more new neurons inside the hidden layers then
neurons inside the hidden layers then
neurons inside the hidden layers then the network specifications speculations
the network specifications speculations
the network specifications speculations can be much more accurate like this so
can be much more accurate like this so
can be much more accurate like this so here by adding more hidden layers with
here by adding more hidden layers with
here by adding more hidden layers with more new ones you have to spend much
more new ones you have to spend much
more new ones you have to spend much more computation power but at the same
more computation power but at the same
more computation power but at the same time the neural networks can compose
time the neural networks can compose
time the neural networks can compose much more complex patterns and extract
much more complex patterns and extract
much more complex patterns and extract the patterns from the large data set how
the patterns from the large data set how
the patterns from the large data set how about this let’s try it this is a data
about this let’s try it this is a data
about this let’s try it this is a data pattern called double spiral if you are
pattern called double spiral if you are
pattern called double spiral if you are a programmer and your director or cast
a programmer and your director or cast
a programmer and your director or cast asked you to cross fight this kind of
asked you to cross fight this kind of
asked you to cross fight this kind of data set what kind of your program
data set what kind of your program
data set what kind of your program called you drive do you want to write
called you drive do you want to write
called you drive do you want to write many issue statements or switch
many issue statements or switch
many issue statements or switch statements with many threshold try
statements with many threshold try
statements with many threshold try checking the x and y partitions no I
checking the x and y partitions no I
checking the x and y partitions no I don’t want to do that instead I would
don’t want to do that instead I would
don’t want to do that instead I would you be using neural networks so that
you be using neural networks so that
you be using neural networks so that neural networks try to looking at the
neural networks try to looking at the
neural networks try to looking at the data points in the data sets to find the
data points in the data sets to find the
data points in the data sets to find the optimal patterns hidden inside the
optimal patterns hidden inside the
optimal patterns hidden inside the training data set that’s right this so
training data set that’s right this so
training data set that’s right this so this is the where in your networks can
this is the where in your networks can
this is the where in your networks can exceed the human performance human
exceed the human performance human
exceed the human performance human programmers performance it can extract
programmers performance it can extract
programmers performance it can extract those hidden patterns inside the
those hidden patterns inside the
those hidden patterns inside the training data set and you don’t have to
training data set and you don’t have to
training data set and you don’t have to specify anything decorative features you
specify anything decorative features you
specify anything decorative features you don’t have to any actual features by
don’t have to any actual features by
don’t have to any actual features by human hand instead computers can find
human hand instead computers can find
human hand instead computers can find the patterns from the training data set
the patterns from the training data set
the patterns from the training data set so that why people are so excited with
so that why people are so excited with
so that why people are so excited with the neural networks and that deep
the neural networks and that deep
the neural networks and that deep learning about this so if you have the
learning about this so if you have the
learning about this so if you have the problems of the identifying handwriting
problems of the identifying handwriting
problems of the identifying handwriting text you can still using the very simple
text you can still using the very simple
text you can still using the very simple GW Express peep equals why kind of
GW Express peep equals why kind of
GW Express peep equals why kind of neural networks to cross find this
neural networks to cross find this
neural networks to cross find this handwriting takes the networks would
handwriting takes the networks would
handwriting takes the networks would come up with disease complex patterns
come up with disease complex patterns
come up with disease complex patterns stacking classify those images into the
stacking classify those images into the
stacking classify those images into the year vectors with the labels like an
year vectors with the labels like an
year vectors with the labels like an eight or seven or six if you want more
eight or seven or six if you want more
eight or seven or six if you want more accuracy than you would have more no
accuracy than you would have more no
accuracy than you would have more no more hidden dangers so that you you
more hidden dangers so that you you
more hidden dangers so that you you could get like a 85 or 95 or 98 like a
could get like a 85 or 95 or 98 like a
could get like a 85 or 95 or 98 like a machine how about this how can you
machine how about this how can you
machine how about this how can you classify these cat images by using
classify these cat images by using
classify these cat images by using neural networks you have to have many
neural networks you have to have many
neural networks you have to have many more layers of neural networks that is
more layers of neural networks that is
more layers of neural networks that is so-called deep neural networks
so-called deep neural networks
so-called deep neural networks this isn’t diagram is called inception
this isn’t diagram is called inception
this isn’t diagram is called inception model pattern 3 there has been published
model pattern 3 there has been published
model pattern 3 there has been published by Google last year where we have used
by Google last year where we have used
by Google last year where we have used 240 hidden layers in a single neural
240 hidden layers in a single neural
240 hidden layers in a single neural network design so it takes much more
network design so it takes much more
network design so it takes much more computation power and time but still we
computation power and time but still we
computation power and time but still we can
can
can much more complex competitions like this
much more complex competitions like this
much more complex competitions like this you know the new ones closer to the
you know the new ones closer to the
you know the new ones closer to the input vector could learn very simple
input vector could learn very simple
input vector could learn very simple pattern rugby’s like you know vertical
pattern rugby’s like you know vertical
pattern rugby’s like you know vertical lines or the horizontal rise but the
lines or the horizontal rise but the
lines or the horizontal rise but the neurons closer to the output vector
neurons closer to the output vector
neurons closer to the output vector could learn much more complex patterns
could learn much more complex patterns
could learn much more complex patterns or compositions such as eyes nose or
or compositions such as eyes nose or
or compositions such as eyes nose or human face again we didn’t put any
human face again we didn’t put any
human face again we didn’t put any features of patterns embedded in the
features of patterns embedded in the
features of patterns embedded in the neural networks before training did so
neural networks before training did so
neural networks before training did so everything can be trained from the data
everything can be trained from the data
everything can be trained from the data set
yeah by using computation power so
yeah by using computation power so that’s the how neural network and deep
that’s the how neural network and deep
that’s the how neural network and deep learning works but as I mentioned it
learning works but as I mentioned it
learning works but as I mentioned it takes so much computation time and
takes so much computation time and
takes so much computation time and training data sets to use the planning’s
training data sets to use the planning’s
training data sets to use the planning’s for the production projects so there are
for the production projects so there are
for the production projects so there are two big challenges right now for the
two big challenges right now for the
two big challenges right now for the users for the deep runnings and this is
users for the deep runnings and this is
users for the deep runnings and this is why keep running is has not been so
why keep running is has not been so
why keep running is has not been so popular for you guys once we have solved
popular for you guys once we have solved
popular for you guys once we have solved these problems raka will have a plenty
these problems raka will have a plenty
these problems raka will have a plenty of computation power with the printed
of computation power with the printed
of computation power with the printed training data set then you can easily
training data set then you can easily
training data set then you can easily apply neural network so deep learning to
apply neural network so deep learning to
apply neural network so deep learning to your existing problems if you a game
your existing problems if you a game
your existing problems if you a game programmer you may want to apply the
programmer you may want to apply the
programmer you may want to apply the deep learnings to analyzing gear you are
deep learnings to analyzing gear you are
deep learnings to analyzing gear you are your game log server logs to check
your game log server logs to check
your game log server logs to check whether a player could be an cheatin
whether a player could be an cheatin
whether a player could be an cheatin player or spammer office weather or if
player or spammer office weather or if
player or spammer office weather or if you web designer or web systems engineer
you web designer or web systems engineer
you web designer or web systems engineer for the ad system then you may want to
for the ad system then you may want to
for the ad system then you may want to apply the logs for the as conversion or
apply the logs for the as conversion or
apply the logs for the as conversion or quick-quick flow rate our logs to neural
quick-quick flow rate our logs to neural
quick-quick flow rate our logs to neural network so that you can get you can have
network so that you can get you can have
network so that you can get you can have computers to learn from the year as log
computers to learn from the year as log
computers to learn from the year as log to get more optimization but you have to
to get more optimization but you have to
to get more optimization but you have to have computation power and training data
have computation power and training data
have computation power and training data so that’s the reason why we have started
so that’s the reason why we have started
so that’s the reason why we have started using Google cloud to train in large
using Google cloud to train in large
using Google cloud to train in large scale neural network Google cloud has
scale neural network Google cloud has
scale neural network Google cloud has racket and
racket and
racket and hundreds of thousand machines in our
hundreds of thousand machines in our
hundreds of thousand machines in our data centers in global and we have been
data centers in global and we have been
data centers in global and we have been building those computers at the data
building those computers at the data
building those computers at the data center as a computer not just a bunch of
center as a computer not just a bunch of
center as a computer not just a bunch of the computers Switzer building we design
the computers Switzer building we design
the computers Switzer building we design each Craster which holds like a ten or
each Craster which holds like a ten or
each Craster which holds like a ten or twenty thousand servers working as a
twenty thousand servers working as a
twenty thousand servers working as a single computer with a multiple our CPUs
single computer with a multiple our CPUs
single computer with a multiple our CPUs so that’s the reason why we can it’s
so that’s the reason why we can it’s
so that’s the reason why we can it’s it’s not so hard for us to deploy try to
it’s not so hard for us to deploy try to
it’s not so hard for us to deploy try to scale neural networks or odds can be
scale neural networks or odds can be
scale neural networks or odds can be created Processing’s to our google cloud
created Processing’s to our google cloud
created Processing’s to our google cloud if there are two basic very fundamental
if there are two basic very fundamental
if there are two basic very fundamental technologies inside google’s that
technologies inside google’s that
technologies inside google’s that supports the data center the computer
supports the data center the computer
supports the data center the computer one is the network we have been building
one is the network we have been building
one is the network we have been building our own hardware support for the network
our own hardware support for the network
our own hardware support for the network switch fabric that is called jupiter
switch fabric that is called jupiter
switch fabric that is called jupiter networks so we union we are not using
networks so we union we are not using
networks so we union we are not using the commercial network switches for most
the commercial network switches for most
the commercial network switches for most cases such as the Cisco or juniper
cases such as the Cisco or juniper
cases such as the Cisco or juniper routers those are not mainstream of our
routers those are not mainstream of our
routers those are not mainstream of our new our network backbones we we have
new our network backbones we we have
new our network backbones we we have been building our own hardware that can
been building our own hardware that can
been building our own hardware that can hold like a hundred thousand pots of 10
hold like a hundred thousand pots of 10
hold like a hundred thousand pots of 10 Gigabit Ethernet ports that can eat at
Gigabit Ethernet ports that can eat at
Gigabit Ethernet ports that can eat at one point to pet a bit per second per
one point to pet a bit per second per
one point to pet a bit per second per our datacenter so that is the networks
our datacenter so that is the networks
our datacenter so that is the networks we have at Google and also contain a
we have at Google and also contain a
we have at Google and also contain a technology called Borg bo is our
technology called Borg bo is our
technology called Borg bo is our proprietary container technologies we
proprietary container technologies we
proprietary container technologies we have been using over 10 years for for
have been using over 10 years for for
have been using over 10 years for for deploying or almost all Google services
deploying or almost all Google services
deploying or almost all Google services such as Google search or Gmail or Google
such as Google search or Gmail or Google
such as Google search or Gmail or Google Maps
Maps
Maps bulk containers account hold up to
bulk containers account hold up to
bulk containers account hold up to 10,000 or 20,000 physical servers in a
10,000 or 20,000 physical servers in a
10,000 or 20,000 physical servers in a single cluster so that you can do the
single cluster so that you can do the
single cluster so that you can do the large scale job scheduling right the
large scale job scheduling right the
large scale job scheduling right the scheduling the CPU cycles or memory
scheduling the CPU cycles or memory
scheduling the CPU cycles or memory spaces or disk i/os with that scale so
spaces or disk i/os with that scale so
spaces or disk i/os with that scale so that reason why you can deploy your
that reason why you can deploy your
that reason why you can deploy your single applications like did the neural
single applications like did the neural
single applications like did the neural network training or the big data
network training or the big data
network training or the big data processing into maybe hundreds or
processing into maybe hundreds or
processing into maybe hundreds or thousands of machines with a single land
thousands of machines with a single land
thousands of machines with a single land of default command
of default command
of default command and Google brain is the project where we
and Google brain is the project where we
and Google brain is the project where we have started applying the Google cloud
have started applying the Google cloud
have started applying the Google cloud technology 2d to build a large-scale
technology 2d to build a large-scale
technology 2d to build a large-scale neural networks this project has started
neural networks this project has started
neural networks this project has started in 2011 and right now the Google brain
in 2011 and right now the Google brain
in 2011 and right now the Google brain has been used for the many many
has been used for the many many
has been used for the many many production project in Google and fast
production project in Google and fast
production project in Google and fast the scalability of Google Google brain
the scalability of Google Google brain
the scalability of Google Google brain project for example rankbrain rankbrain
project for example rankbrain rankbrain
project for example rankbrain rankbrain is our GU algorithms we are using for
is our GU algorithms we are using for
is our GU algorithms we are using for the ranking of Google search service
the ranking of Google search service
the ranking of Google search service right now since last year that has been
right now since last year that has been
right now since last year that has been using Google Google brain infrastructure
using Google Google brain infrastructure
using Google Google brain infrastructure and with five hundred nodes and that can
and with five hundred nodes and that can
and with five hundred nodes and that can perform at three hundred times faster
perform at three hundred times faster
perform at three hundred times faster than single node so that means if you
than single node so that means if you
than single node so that means if you are training your deep learning model
are training your deep learning model
are training your deep learning model with single servers then you would take
with single servers then you would take
with single servers then you would take 300 times longer than Google engineers
300 times longer than Google engineers
300 times longer than Google engineers and inception is the model for the
and inception is the model for the
and inception is the model for the visual visual recognition we can use 50
visual visual recognition we can use 50
visual visual recognition we can use 50 GPUs to accelerate the performance at 40
GPUs to accelerate the performance at 40
GPUs to accelerate the performance at 40 times faster so that reason those are
times faster so that reason those are
times faster so that reason those are the reason why Google has been so strong
the reason why Google has been so strong
the reason why Google has been so strong on applying deep running for the
on applying deep running for the
on applying deep running for the production project such as the alphago
production project such as the alphago
production project such as the alphago Frady we have the series of the core
Frady we have the series of the core
Frady we have the series of the core matches with the core professional they
matches with the core professional they
matches with the core professional they have been using the Google brain
have been using the Google brain
have been using the Google brain infrastructure for the training as well
infrastructure for the training as well
infrastructure for the training as well as the prediction of the Google match
as the prediction of the Google match
as the prediction of the Google match Google search has been using deep with
Google search has been using deep with
Google search has been using deep with the brain of since last year and we have
the brain of since last year and we have
the brain of since last year and we have been using the machine learning
been using the machine learning
been using the machine learning technologies for the optimizing the data
technologies for the optimizing the data
technologies for the optimizing the data center operation and also or she our
center operation and also or she our
center operation and also or she our natural language processing and visual
natural language processing and visual
natural language processing and visual recognition of speech recognition such
recognition of speech recognition such
recognition of speech recognition such as the Google photos what he voiced the
as the Google photos what he voiced the
as the Google photos what he voiced the conventions of the androids we have over
conventions of the androids we have over
conventions of the androids we have over 60 production projects that has been
60 production projects that has been
60 production projects that has been using Google brain and deep learnings
using Google brain and deep learnings
using Google brain and deep learnings for last a couple of years now we have
for last a couple of years now we have
for last a couple of years now we have started to externalizing this power of
started to externalizing this power of
started to externalizing this power of Google brain to external developers
the first product is called crowd vision
the first product is called crowd vision API and the second product is called
API and the second product is called
API and the second product is called crowd speech API crowd vision API is an
crowd speech API crowd vision API is an
crowd speech API crowd vision API is an image analysis IPA that provides the
image analysis IPA that provides the
image analysis IPA that provides the pre-trained model so you don’t have to
pre-trained model so you don’t have to
pre-trained model so you don’t have to train your own neural network and you
train your own neural network and you
train your own neural network and you also don’t have to have the any skill
also don’t have to have the any skill
also don’t have to have the any skill set for the machine learning so it’s
set for the machine learning so it’s
set for the machine learning so it’s just on REST API you can just upload
just on REST API you can just upload
just on REST API you can just upload your photo image to API then you repeat
your photo image to API then you repeat
your photo image to API then you repeat receiving JSON result in a few seconds
receiving JSON result in a few seconds
receiving JSON result in a few seconds there has the the analysis result and
there has the the analysis result and
there has the the analysis result and it’s free to start trying out up to
it’s free to start trying out up to
it’s free to start trying out up to 1,000 images per month and it’s general
1,000 images per month and it’s general
1,000 images per month and it’s general generally available right now so it’s
generally available right now so it’s
generally available right now so it’s ready to be used for the production
ready to be used for the production
ready to be used for the production project it has six different features to
project it has six different features to
project it has six different features to be detected labial detections means that
be detected labial detections means that
be detected labial detections means that you can put any labels or categories on
you can put any labels or categories on
you can put any labels or categories on any images you uploaded for example if
any images you uploaded for example if
any images you uploaded for example if you uploading the cat images then the
you uploading the cat images then the
you uploading the cat images then the API will be returning the Arabians such
API will be returning the Arabians such
API will be returning the Arabians such as a cat or pet face detections can
as a cat or pet face detections can
as a cat or pet face detections can detect the location of face in the image
detect the location of face in the image
detect the location of face in the image OCR I can convert the text on image to a
OCR I can convert the text on image to a
OCR I can convert the text on image to a string explicit content detection means
string explicit content detection means
string explicit content detection means that you can check whether the images
that you can check whether the images
that you can check whether the images can contain the images contain the the
can contain the images contain the the
can contain the images contain the the adult or violent images on our landmark
adult or violent images on our landmark
adult or violent images on our landmark detection can detect the location of the
detection can detect the location of the
detection can detect the location of the images or popular places and you can
images or popular places and you can
images or popular places and you can also detect the product or corporate
also detect the product or corporate
also detect the product or corporate role let’s take a look at the
role let’s take a look at the
role let’s take a look at the demonstration
so I’d like to show a demonstration by
so I’d like to show a demonstration by video at first this is the
video at first this is the
video at first this is the demonstrations by using the Raspberry Pi
demonstrations by using the Raspberry Pi
demonstrations by using the Raspberry Pi robot that sends the image to division
robot that sends the image to division
robot that sends the image to division API cloud vision provides powerful image
API cloud vision provides powerful image
API cloud vision provides powerful image analytics capabilities as easy to use
analytics capabilities as easy to use
analytics capabilities as easy to use api’s it enables application developers
api’s it enables application developers
api’s it enables application developers to build the next generation of
to build the next generation of
to build the next generation of application that can see and understand
application that can see and understand
application that can see and understand the content within the images the
the content within the images the
the content within the images the service is built on powerful computer
service is built on powerful computer
service is built on powerful computer vision models that power several to firm
vision models that power several to firm
vision models that power several to firm Google services the service enables
Google services the service enables
Google services the service enables developers to detect a broad set of
developers to detect a broad set of
developers to detect a broad set of entities within an image from everyday
entities within an image from everyday
entities within an image from everyday objects to faces in product logos the
objects to faces in product logos the
objects to faces in product logos the service is so easy to use as one example
service is so easy to use as one example
service is so easy to use as one example of the use cases you can have any
of the use cases you can have any
of the use cases you can have any Raspberry Pi robot like gulp I go
Raspberry Pi robot like gulp I go
Raspberry Pi robot like gulp I go calling the cloud vision API directly so
calling the cloud vision API directly so
calling the cloud vision API directly so the broad can sum the images taken by
the broad can sum the images taken by
the broad can sum the images taken by its camera to the cloud and can get the
its camera to the cloud and can get the
its camera to the cloud and can get the analysis results in real time it detects
analysis results in real time it detects
analysis results in real time it detects faces in the image along with the
faces in the image along with the
faces in the image along with the associated emotions the cloud vision API
associated emotions the cloud vision API
associated emotions the cloud vision API is also able to detect entities within
is also able to detect entities within
is also able to detect entities within the image now let’s see how facial
the image now let’s see how facial
the image now let’s see how facial detection works cloud vision detect
detection works cloud vision detect
detection works cloud vision detect spaces on the picture and returns the
spaces on the picture and returns the
spaces on the picture and returns the positions of eyes nose and mouth so you
positions of eyes nose and mouth so you
positions of eyes nose and mouth so you can program the bot to follow the face
it also detects emotions such as joy
it also detects emotions such as joy anger surprise and sorrow so the bottom
anger surprise and sorrow so the bottom
anger surprise and sorrow so the bottom moved toward smiling faces or avoid
moved toward smiling faces or avoid
moved toward smiling faces or avoid anger or surprise face one of the very
anger or surprise face one of the very
anger or surprise face one of the very interesting features of cloud vision API
interesting features of cloud vision API
interesting features of cloud vision API is the entity detection that means it
is the entity detection that means it
is the entity detection that means it detects any objects you like you see
detects any objects you like you see
detects any objects you like you see cloud visitors likes developers to take
cloud visitors likes developers to take
cloud visitors likes developers to take advantage of Google’s latest machine
advantage of Google’s latest machine
advantage of Google’s latest machine learning technologies quite easily
learning technologies quite easily
learning technologies quite easily please go to cloud.google.com slash
please go to cloud.google.com slash
please go to cloud.google.com slash vision to learn more and I have another
vision to learn more and I have another
vision to learn more and I have another interesting demonstrations that is made
interesting demonstrations that is made
interesting demonstrations that is made by using the vision API that if this is
by using the vision API that if this is
by using the vision API that if this is called vision Explorer demonstrations
called vision Explorer demonstrations
called vision Explorer demonstrations where we have imported 80,000 images
where we have imported 80,000 images
where we have imported 80,000 images downloaded from Wikimedia Commons and
downloaded from Wikimedia Commons and
downloaded from Wikimedia Commons and uploaded to the Google Cloud storage and
uploaded to the Google Cloud storage and
uploaded to the Google Cloud storage and applied the vision API analysis so here
applied the vision API analysis so here
applied the vision API analysis so here we have the cluster of the images it is
we have the cluster of the images it is
we have the cluster of the images it is 80 thousand images and each cluster has
80 thousand images and each cluster has
80 thousand images and each cluster has the labels such as snow or transport
the labels such as snow or transport
the labels such as snow or transport residential area means that the the
residential area means that the the
residential area means that the the cluster of the similar images for
cluster of the similar images for
cluster of the similar images for example if you take a look at here let’s
example if you take a look at here let’s
example if you take a look at here let’s go to a plant so each single dot
go to a plant so each single dot
go to a plant so each single dot represents your thumbnail of the
represents your thumbnail of the
represents your thumbnail of the uploaded images so if you go to the
uploaded images so if you go to the
uploaded images so if you go to the plant
plant
plant Craster there must be some cluster of
Craster there must be some cluster of
Craster there must be some cluster of the oh it’s oh it’s not showing why let
the oh it’s oh it’s not showing why let
the oh it’s oh it’s not showing why let me redraw this maybe because I’m using
me redraw this maybe because I’m using
me redraw this maybe because I’m using tethering
okay let’s go directly to the cat
okay let’s go directly to the cat cluster so in this cluster we have many
cluster so in this cluster we have many
cluster so in this cluster we have many many cats and closer to the cat cora’s
many cats and closer to the cat cora’s
many cats and closer to the cat cora’s we have the crust for dogs let’s go back
we have the crust for dogs let’s go back
we have the crust for dogs let’s go back to the cat cluster and if you click to
to the cat cluster and if you click to
to the cat cluster and if you click to image thumbnail image and you’ll be
image thumbnail image and you’ll be
image thumbnail image and you’ll be seeing the analysis result from the API
seeing the analysis result from the API
seeing the analysis result from the API right this the API thinks this must be a
right this the API thinks this must be a
right this the API thinks this must be a mimic of cat and it’s a cat as a pet or
mimic of cat and it’s a cat as a pet or
mimic of cat and it’s a cat as a pet or it must it must be a British Shorthair
it must it must be a British Shorthair
it must it must be a British Shorthair so this is these some things you can do
so this is these some things you can do
so this is these some things you can do with the deep learning technology and
with the deep learning technology and
with the deep learning technology and those results are returned in a JSON
those results are returned in a JSON
those results are returned in a JSON format erectus and with statistical
format erectus and with statistical
format erectus and with statistical stretches we can show it in a GUI if
stretches we can show it in a GUI if
stretches we can show it in a GUI if image contains any text inside it then
image contains any text inside it then
image contains any text inside it then we can come back convert it into the
we can come back convert it into the
we can come back convert it into the extreme example with this you can have
extreme example with this you can have
extreme example with this you can have the string like these three Kangaroos
the string like these three Kangaroos
the string like these three Kangaroos crossing next to you images if images
crossing next to you images if images
crossing next to you images if images contains faces also this API doesn’t
contains faces also this API doesn’t
contains faces also this API doesn’t support any personal identification or
support any personal identification or
support any personal identification or the personal recognition but it can
the personal recognition but it can
the personal recognition but it can detect the location of the faces with
detect the location of the faces with
detect the location of the faces with landmark locations such as nose and
landmark locations such as nose and
landmark locations such as nose and mouth and also it can recognize the
mouth and also it can recognize the
mouth and also it can recognize the emotions such as joy sorrow and anger
emotions such as joy sorrow and anger
emotions such as joy sorrow and anger and surprising in this confidence level
and surprising in this confidence level
and surprising in this confidence level and if your picture contains any popular
and if your picture contains any popular
and if your picture contains any popular places such as this dinner API can
places such as this dinner API can
places such as this dinner API can return the name of the landmark the API
return the name of the landmark the API
return the name of the landmark the API thinks it must be an image of the Citi
thinks it must be an image of the Citi
thinks it must be an image of the Citi Field Stadium in New York City with the
Field Stadium in New York City with the
Field Stadium in New York City with the longitude and latitude so you can easily
longitude and latitude so you can easily
longitude and latitude so you can easily in a put a marker on the Google Maps
in a put a marker on the Google Maps
in a put a marker on the Google Maps it’s too slow so I’m cutting it off also
it’s too slow so I’m cutting it off also
it’s too slow so I’m cutting it off also you can detect the product and corporate
you can detect the product and corporate
you can detect the product and corporate robot
like this Android so this was the this
like this Android so this was the this vision API so it’s ready to be used for
vision API so it’s ready to be used for
vision API so it’s ready to be used for any applications and another API is
any applications and another API is
any applications and another API is called speech API which also provides
called speech API which also provides
called speech API which also provides the pre-trained model for the voice
the pre-trained model for the voice
the pre-trained model for the voice recognition so you don’t have to have
recognition so you don’t have to have
recognition so you don’t have to have any skill set or experiments with the
any skill set or experiments with the
any skill set or experiments with the voice recognition or training neural
voice recognition or training neural
voice recognition or training neural networks for doing doing that it’s just
networks for doing doing that it’s just
networks for doing doing that it’s just on REST API and G RPC API so you can
on REST API and G RPC API so you can
on REST API and G RPC API so you can just upload your audio data to the API
just upload your audio data to the API
just upload your audio data to the API and you’ll be receiving a result in the
and you’ll be receiving a result in the
and you’ll be receiving a result in the few seconds it supports over 80
few seconds it supports over 80
few seconds it supports over 80 languages and dialects it supports both
languages and dialects it supports both
languages and dialects it supports both real-time recognition and battery
real-time recognition and battery
real-time recognition and battery recognition the API is still in limited
recognition the API is still in limited
recognition the API is still in limited preview so if you go to the speech
preview so if you go to the speech
preview so if you go to the speech cloud.google.com speech then you have to
cloud.google.com speech then you have to
cloud.google.com speech then you have to sign up with the form for immediate
sign up with the form for immediate
sign up with the form for immediate limited preview access but we hope to
limited preview access but we hope to
limited preview access but we hope to make it public better maybe in a couple
make it public better maybe in a couple
make it public better maybe in a couple of weeks I suppose let’s show some
of weeks I suppose let’s show some
of weeks I suppose let’s show some demonstration I’m not sure if this works
demonstration I’m not sure if this works
demonstration I’m not sure if this works in the event or not because this is the
in the event or not because this is the
in the event or not because this is the first time to try this and and I have
first time to try this and and I have
first time to try this and and I have some accent problems so I’m not sure I
some accent problems so I’m not sure I
some accent problems so I’m not sure I really not sure if this works or not but
really not sure if this works or not but
really not sure if this works or not but best right hello this is a testing of I
best right hello this is a testing of I
best right hello this is a testing of I think it’s not working
hello this is a test of voice
hello this is a test of voice recognition bar by Google Cloud machine
recognition bar by Google Cloud machine
recognition bar by Google Cloud machine learning oh yeah
learning oh yeah
learning oh yeah so final result is you know not bad
so final result is you know not bad
so final result is you know not bad right and you also saw the fast response
right and you also saw the fast response
right and you also saw the fast response so you could get the recognition result
so you could get the recognition result
so you could get the recognition result in recent one second district Iraq a 0.5
in recent one second district Iraq a 0.5
in recent one second district Iraq a 0.5 seconds in real time so those are the
seconds in real time so those are the
seconds in real time so those are the api’s and but those api’s here are
api’s and but those api’s here are
api’s and but those api’s here are pre-trained model so that means you
pre-trained model so that means you
pre-trained model so that means you cannot train your own model with those
cannot train your own model with those
cannot train your own model with those aps and well one would if we country
aps and well one would if we country
aps and well one would if we country asked questions for those api is that
asked questions for those api is that
asked questions for those api is that whether the google will be you know
whether the google will be you know
whether the google will be you know looking at the uploaded images or the
looking at the uploaded images or the
looking at the uploaded images or the audio data to train your own model or
audio data to train your own model or
audio data to train your own model or doing some more research and as know
doing some more research and as know
doing some more research and as know those api saudi all of our products are
those api saudi all of our products are
those api saudi all of our products are provided by the Google cloud platform is
provided by the Google cloud platform is
provided by the Google cloud platform is is under the terms and conditions of DCP
is under the terms and conditions of DCP
is under the terms and conditions of DCP that has here some sections for the
that has here some sections for the
that has here some sections for the customer customer data we don’t look at
customer customer data we don’t look at
customer customer data we don’t look at the customer data except for the very
the customer data except for the very
the customer data except for the very special cases for the troubleshooting or
special cases for the troubleshooting or
special cases for the troubleshooting or emergency situation so basically we
emergency situation so basically we
emergency situation so basically we don’t look at the gyro data uploaded to
don’t look at the gyro data uploaded to
don’t look at the gyro data uploaded to the cloud but at the same time so you
the cloud but at the same time so you
the cloud but at the same time so you can so the APS cannot train cannot do
can so the APS cannot train cannot do
can so the APS cannot train cannot do the trainings for your data or your
the trainings for your data or your
the trainings for your data or your applications so that’s the reason why we
applications so that’s the reason why we
applications so that’s the reason why we provide the other options for machine
provide the other options for machine
provide the other options for machine learning with a stencil or cloud machine
learning with a stencil or cloud machine
learning with a stencil or cloud machine learning the other d-dick frameworks and
learning the other d-dick frameworks and
learning the other d-dick frameworks and platforms that can used for train your
platforms that can used for train your
platforms that can used for train your own data set train your own machine
own data set train your own machine
own data set train your own machine learning and neural network what is 10
learning and neural network what is 10
learning and neural network what is 10 so for tensile Pro is an open-source
so for tensile Pro is an open-source
so for tensile Pro is an open-source driver of your machine intelligence we
driver of your machine intelligence we
driver of your machine intelligence we have published the libraries last
have published the libraries last
have published the libraries last November and this is the the
November and this is the the
November and this is the the actual framework we are right now using
actual framework we are right now using
actual framework we are right now using us via Google research of Google brain
us via Google research of Google brain
us via Google research of Google brain team so it’s not something’s
team so it’s not something’s
team so it’s not something’s outdated or stay out since the latest
outdated or stay out since the latest
outdated or stay out since the latest machine learning framework we are using
machine learning framework we are using
machine learning framework we are using are right now at Google for example if
are right now at Google for example if
are right now at Google for example if you want to design this
you want to design this
you want to design this restaurant work swag DW Express Pico why
restaurant work swag DW Express Pico why
restaurant work swag DW Express Pico why you can use Python to write it in a
you can use Python to write it in a
you can use Python to write it in a single line of code Roxas you can put
single line of code Roxas you can put
single line of code Roxas you can put the image of cat here this vector and
the image of cat here this vector and
the image of cat here this vector and then you would have an output vector
then you would have an output vector
then you would have an output vector that represents the labels of the
that represents the labels of the
that represents the labels of the detected objects like a cat or human
detected objects like a cat or human
detected objects like a cat or human face and you can let computers to find
face and you can let computers to find
face and you can let computers to find the obits and biases so it’s so simple
the obits and biases so it’s so simple
the obits and biases so it’s so simple and also it’s really simple to train
and also it’s really simple to train
and also it’s really simple to train your networks because you can just write
your networks because you can just write
your networks because you can just write this single line to have your networks
this single line to have your networks
this single line to have your networks trained for your training data set by
trained for your training data set by
trained for your training data set by using by specifying the algorithm
using by specifying the algorithm
using by specifying the algorithm Stryker gradient is set you don’t have
Stryker gradient is set you don’t have
Stryker gradient is set you don’t have to implement your own the procedural
to implement your own the procedural
to implement your own the procedural code or called – implementing the each
code or called – implementing the each
code or called – implementing the each the optimization logic actually I’m not
the optimization logic actually I’m not
the optimization logic actually I’m not good at math or those machine learning
good at math or those machine learning
good at math or those machine learning algorithms but still I can just copy and
algorithms but still I can just copy and
algorithms but still I can just copy and paste the sample code to my laptop and
paste the sample code to my laptop and
paste the sample code to my laptop and I’m praying with my own data sets missed
I’m praying with my own data sets missed
I’m praying with my own data sets missed and so forth so you can just let the
and so forth so you can just let the
and so forth so you can just let the chancel for runtimes to do the
chancel for runtimes to do the
chancel for runtimes to do the optimization and also the tool provides
optimization and also the tool provides
optimization and also the tool provides you a very good visualization tool so
you a very good visualization tool so
you a very good visualization tool so one of the problems we had at Google for
one of the problems we had at Google for
one of the problems we had at Google for applying the neural networks to the
applying the neural networks to the
applying the neural networks to the production production problem is the
production production problem is the
production production problem is the debugging so if you have many more
debugging so if you have many more
debugging so if you have many more hidden layers inside the neural networks
hidden layers inside the neural networks
hidden layers inside the neural networks you have to check the all these stages
you have to check the all these stages
you have to check the all these stages of the parameters whether the parameters
of the parameters whether the parameters
of the parameters whether the parameters are converging in a right direction or
are converging in a right direction or
are converging in a right direction or the parameters could be you know going
the parameters could be you know going
the parameters could be you know going away and having a wrong number such as a
away and having a wrong number such as a
away and having a wrong number such as a na or 0 elsewhere so it’s really
na or 0 elsewhere so it’s really
na or 0 elsewhere so it’s really important to visualize what’s happening
important to visualize what’s happening
important to visualize what’s happening inside in your networks and tensorflow
inside in your networks and tensorflow
inside in your networks and tensorflow provides the tool and also the
provides the tool and also the
provides the tool and also the portability is another important aspect
portability is another important aspect
portability is another important aspect of the framework so once you have
of the framework so once you have
of the framework so once you have defined your neural networks with
defined your neural networks with
defined your neural networks with Python code of tensorflow then you can
Python code of tensorflow then you can
Python code of tensorflow then you can start running you and you ready to work
start running you and you ready to work
start running you and you ready to work training or prediction with your laptop
training or prediction with your laptop
training or prediction with your laptop like a Mac or Windows Raptor but you
like a Mac or Windows Raptor but you
like a Mac or Windows Raptor but you will find that your laptop is too slow
will find that your laptop is too slow
will find that your laptop is too slow to trying the trendy deep neural
to trying the trendy deep neural
to trying the trendy deep neural networks so maybe you may soon want to
networks so maybe you may soon want to
networks so maybe you may soon want to buy some GPU class and instead of maybe
buy some GPU class and instead of maybe
buy some GPU class and instead of maybe 2 or 3 GB because in a single box but
2 or 3 GB because in a single box but
2 or 3 GB because in a single box but still usually it takes like a few few
still usually it takes like a few few
still usually it takes like a few few days usually a few days or maybe some
days usually a few days or maybe some
days usually a few days or maybe some people spending a few weeks to to do the
people spending a few weeks to to do the
people spending a few weeks to to do the trainings on their neural networks so it
trainings on their neural networks so it
trainings on their neural networks so it takes so much computation time so in
takes so much computation time so in
takes so much computation time so in that case you can applaud your tensor
that case you can applaud your tensor
that case you can applaud your tensor flow graph to Google cloud so that you
flow graph to Google cloud so that you
flow graph to Google cloud so that you can utilize the power of the tents or
can utilize the power of the tents or
can utilize the power of the tents or maybe hundreds of GPU instances we have
maybe hundreds of GPU instances we have
maybe hundreds of GPU instances we have we’re running at Google cloud and also
we’re running at Google cloud and also
we’re running at Google cloud and also once you have finished your training
once you have finished your training
once you have finished your training then the size of the parameter sets
then the size of the parameter sets
then the size of the parameter sets could be fit into our hundreds of
could be fit into our hundreds of
could be fit into our hundreds of megabytes or tens of megabytes then you
megabytes or tens of megabytes then you
megabytes or tens of megabytes then you can easily copy that parameter sets into
can easily copy that parameter sets into
can easily copy that parameter sets into the smaller devices mobile devices or
the smaller devices mobile devices or
the smaller devices mobile devices or IOT devices such as Android iOS or maybe
IOT devices such as Android iOS or maybe
IOT devices such as Android iOS or maybe Raspberry Pi so that you know you can
Raspberry Pi so that you know you can
Raspberry Pi so that you know you can have those devices doing the prediction
have those devices doing the prediction
have those devices doing the prediction like image recognition or voice
like image recognition or voice
like image recognition or voice retention without using any internet
retention without using any internet
retention without using any internet connection
connection
connection everything could be implemented within
everything could be implemented within
everything could be implemented within the framework of tensorflow and with the
the framework of tensorflow and with the
the framework of tensorflow and with the at the last Google i/o it was about 1
at the last Google i/o it was about 1
at the last Google i/o it was about 1 months ago we have announced a new
months ago we have announced a new
months ago we have announced a new technology called tensor processing unit
technology called tensor processing unit
technology called tensor processing unit this is a replacement not a replacement
this is a replacement not a replacement
this is a replacement not a replacement maybe a complementary technology for the
maybe a complementary technology for the
maybe a complementary technology for the GPU and CPU so so far or maybe right now
GPU and CPU so so far or maybe right now
GPU and CPU so so far or maybe right now the any deep neural networks researchers
the any deep neural networks researchers
the any deep neural networks researchers or developers outside Google is using
or developers outside Google is using
or developers outside Google is using GPUs mostly for training the neural
GPUs mostly for training the neural
GPUs mostly for training the neural networks because it’s a matrix
networks because it’s a matrix
networks because it’s a matrix operations and by using GPUs you can
operations and by using GPUs you can
operations and by using GPUs you can accelerate the matrix office
accelerate the matrix office
accelerate the matrix office ten times or maybe 40 times faster so
ten times or maybe 40 times faster so
ten times or maybe 40 times faster so that’s what typical neural networks
that’s what typical neural networks
that’s what typical neural networks users are doing right now but the
users are doing right now but the
users are doing right now but the devices problem for GPU is the power
devices problem for GPU is the power
devices problem for GPU is the power consumption each consumers record 100
consumption each consumers record 100
consumption each consumers record 100 watts or 200 watts per GPU card and we
watts or 200 watts per GPU card and we
watts or 200 watts per GPU card and we are having we were using thousands of
are having we were using thousands of
are having we were using thousands of them in a Google Data Center and power
them in a Google Data Center and power
them in a Google Data Center and power Concepcion is becoming the Rogers
Concepcion is becoming the Rogers
Concepcion is becoming the Rogers problem so by by designing the Asics or
problem so by by designing the Asics or
problem so by by designing the Asics or the editorship specifically for the
the editorship specifically for the
the editorship specifically for the tensor flow or deep neural networks we
tensor flow or deep neural networks we
tensor flow or deep neural networks we were able to reduce the power
were able to reduce the power
were able to reduce the power consumption and gain the ten times
consumption and gain the ten times
consumption and gain the ten times better for our performance – powerful
better for our performance – powerful
better for our performance – powerful performance for what result and we also
performance for what result and we also
performance for what result and we also use the special techniques such as the
use the special techniques such as the
use the special techniques such as the bit quantization rather than using a
bit quantization rather than using a
bit quantization rather than using a 32-bit or 16-bit to calculate everything
32-bit or 16-bit to calculate everything
32-bit or 16-bit to calculate everything every matrix operations we use the
every matrix operations we use the
every matrix operations we use the quantization strike a quantized into the
quantization strike a quantized into the
quantization strike a quantized into the 8-bit where there’s not so not so much
8-bit where there’s not so not so much
8-bit where there’s not so not so much loss of the accuracy so that you can fit
loss of the accuracy so that you can fit
loss of the accuracy so that you can fit much bigger parameters into a very small
much bigger parameters into a very small
much bigger parameters into a very small memory footprint and we have been using
memory footprint and we have been using
memory footprint and we have been using GTP’s for many production projects
GTP’s for many production projects
GTP’s for many production projects already run greying alphago and google
already run greying alphago and google
already run greying alphago and google photos speech recognitions these are all
photos speech recognitions these are all
photos speech recognitions these are all has been using tepees since a couple
has been using tepees since a couple
has been using tepees since a couple months actually we have been using GPS
months actually we have been using GPS
months actually we have been using GPS for less than one year and the if you
for less than one year and the if you
for less than one year and the if you want to yeah we have been I haven’t
want to yeah we have been I haven’t
want to yeah we have been I haven’t discussed describing about the power of
discussed describing about the power of
discussed describing about the power of the Google brain such as number of the
the Google brain such as number of the
the Google brain such as number of the CPUs GPUs and TP is and if you want to
CPUs GPUs and TP is and if you want to
CPUs GPUs and TP is and if you want to utilize the power of Google brand
utilize the power of Google brand
utilize the power of Google brand infrastructure here’s the product we
infrastructure here’s the product we
infrastructure here’s the product we provide which is called cloud machine
provide which is called cloud machine
provide which is called cloud machine learning
learning
learning crud machine learning is a fully managed
crud machine learning is a fully managed
crud machine learning is a fully managed distributed training environment for
distributed training environment for
distributed training environment for your tester for graph so once you have
your tester for graph so once you have
your tester for graph so once you have written
written
written wrote your tensor flow graph and run run
wrote your tensor flow graph and run run
wrote your tensor flow graph and run run it on the laptop then you can upload the
it on the laptop then you can upload the
it on the laptop then you can upload the same types of rock for graph to Google
same types of rock for graph to Google
same types of rock for graph to Google Cloud messing learning so that you can
Cloud messing learning so that you can
Cloud messing learning so that you can specify the number of the GPS you want
specify the number of the GPS you want
specify the number of the GPS you want to use with the service suggested you
to use with the service suggested you
to use with the service suggested you such as 20 nodes or 50 notes to do the
such as 20 nodes or 50 notes to do the
such as 20 nodes or 50 notes to do the acceleration only training when a crowd
acceleration only training when a crowd
acceleration only training when a crowd Mao is in the limited preview so you
Mao is in the limited preview so you
Mao is in the limited preview so you have to sign up to start trying out but
have to sign up to start trying out but
have to sign up to start trying out but maybe I we suppose that the for
maybe I we suppose that the for
maybe I we suppose that the for availability record public better will
availability record public better will
availability record public better will be sometimes later in this year if you
be sometimes later in this year if you
be sometimes later in this year if you go to the YouTube then you can take a
go to the YouTube then you can take a
go to the YouTube then you can take a look at the actual demonstration of
look at the actual demonstration of
look at the actual demonstration of kratom a budget of teens where they have
kratom a budget of teens where they have
kratom a budget of teens where they have where he has presented demonstrated the
where he has presented demonstrated the
where he has presented demonstrated the actual chance of grotessa fraud based
actual chance of grotessa fraud based
actual chance of grotessa fraud based neural networks that takes 8 hours with
neural networks that takes 8 hours with
neural networks that takes 8 hours with single node but if you upload the same
single node but if you upload the same
single node but if you upload the same tensor flow graph to the crowd email
tensor flow graph to the crowd email
tensor flow graph to the crowd email then you can accelerate the performance
then you can accelerate the performance
then you can accelerate the performance to up to 15 times faster that means you
to up to 15 times faster that means you
to up to 15 times faster that means you could get the result of the trainings
could get the result of the trainings
could get the result of the trainings within 30 minutes rather than waiting
within 30 minutes rather than waiting
within 30 minutes rather than waiting for the 8 hours that is the speed we are
for the 8 hours that is the speed we are
for the 8 hours that is the speed we are seeing inside Google for any deep
seeing inside Google for any deep
seeing inside Google for any deep learning deployment and we externalizing
learning deployment and we externalizing
learning deployment and we externalizing the dispo to to you guys to have you
the dispo to to you guys to have you
the dispo to to you guys to have you utilizing the power for 40 you are for
utilizing the power for 40 you are for
utilizing the power for 40 you are for solving your own problems and also
solving your own problems and also
solving your own problems and also claudemir can be used for the production
claudemir can be used for the production
claudemir can be used for the production as well not only for the training and
as well not only for the training and
as well not only for the training and indeed mo solutions he presented at he
indeed mo solutions he presented at he
indeed mo solutions he presented at he demonstrated that the crowd email could
demonstrated that the crowd email could
demonstrated that the crowd email could be used for the body predictions at 300
be used for the body predictions at 300
be used for the body predictions at 300 meters per second so those are the the
meters per second so those are the the
meters per second so those are the the topics I have covered and now we have
topics I have covered and now we have
topics I have covered and now we have two different products one is the Amero
two different products one is the Amero
two different products one is the Amero api’s like a vision API or speech API
api’s like a vision API or speech API
api’s like a vision API or speech API where you can just upload your own data
where you can just upload your own data
where you can just upload your own data to cloud so that you will be getting the
to cloud so that you will be getting the
to cloud so that you will be getting the results in a few seconds and if you want
results in a few seconds and if you want
results in a few seconds and if you want to train your own neural networks
to train your own neural networks
to train your own neural networks then you can use a try using the
then you can use a try using the
then you can use a try using the intensive roll or crud machine learning
intensive roll or crud machine learning
intensive roll or crud machine learning so that you can accelerate training
so that you can accelerate training
so that you can accelerate training you’re on or you’re on your network so
you’re on or you’re on your network so
you’re on or you’re on your network so if you take a look at the links on the
if you take a look at the links on the
if you take a look at the links on the resources of this lesson that you can
resources of this lesson that you can
resources of this lesson that you can start trying out those products right
start trying out those products right
start trying out those products right now thank you so much yeah I got two
now thank you so much yeah I got two
now thank you so much yeah I got two questions Trulia
questions Trulia
questions Trulia first of all few steps back on the
first of all few steps back on the
first of all few steps back on the machine learning and neural networks
machine learning and neural networks
machine learning and neural networks mm-hmm
mm-hmm
mm-hmm you talked about more hidden layers to
you talked about more hidden layers to
you talked about more hidden layers to more complex algorithm yeah but Sturm
more complex algorithm yeah but Sturm
more complex algorithm yeah but Sturm x-fighters they’re like we were told in
x-fighters they’re like we were told in
x-fighters they’re like we were told in University to use yeah you’re feeling to
University to use yeah you’re feeling to
University to use yeah you’re feeling to see how many layers you have any tips on
see how many layers you have any tips on
see how many layers you have any tips on that or yeah yeah that’s actually a
that or yeah yeah that’s actually a
that or yeah yeah that’s actually a really good question
really good question
really good question so maybe question is is there any good
so maybe question is is there any good
so maybe question is is there any good practice on designing the Union neural
practice on designing the Union neural
practice on designing the Union neural networks right so as far as I know
networks right so as far as I know
networks right so as far as I know there’s no one theory to optimize your
there’s no one theory to optimize your
there’s no one theory to optimize your design of neural networks so everybody
design of neural networks so everybody
design of neural networks so everybody even in the fifth at Google you know
even in the fifth at Google you know
even in the fifth at Google you know when I asked the Google research team
when I asked the Google research team
when I asked the Google research team and people would say you know let’s
and people would say you know let’s
and people would say you know let’s start with the five 5 children
start with the five 5 children
start with the five 5 children let’s see how it works all right so
let’s see how it works all right so
let’s see how it works all right so strata and that’s the largest challenge
strata and that’s the largest challenge
strata and that’s the largest challenge we have right now for deploying the
we have right now for deploying the
we have right now for deploying the neural networks for your own data or
neural networks for your own data or
neural networks for your own data or applications so you have to do it in
applications so you have to do it in
applications so you have to do it in many many trials just right the the
many many trials just right the the
many many trials just right the the people in the pharmaceutical companies
people in the pharmaceutical companies
people in the pharmaceutical companies trying to create a new drug so you have
trying to create a new drug so you have
trying to create a new drug so you have to have a different combination of
to have a different combination of
to have a different combination of obably hyperparameters hyper parameters
obably hyperparameters hyper parameters
obably hyperparameters hyper parameters means the parameters such as the number
means the parameters such as the number
means the parameters such as the number of hidden layers or the or new ones or
of hidden layers or the or new ones or
of hidden layers or the or new ones or the way you can import the data or
the way you can import the data or
the way you can import the data or extracting features so you have to try
extracting features so you have to try
extracting features so you have to try out every different combinations that’s
out every different combinations that’s
out every different combinations that’s the problem
the problem
the problem and also it takes much much computation
and also it takes much much computation
and also it takes much much computation power yeah I think that’s what all do
power yeah I think that’s what all do
power yeah I think that’s what all do yeah it’s not a theory behind it and a
yeah it’s not a theory behind it and a
yeah it’s not a theory behind it and a little bit later in the presentation but
little bit later in the presentation but
little bit later in the presentation but use cases will be supported in the near
use cases will be supported in the near
use cases will be supported in the near future
future
future we’re going from post training to
we’re going from post training to
we’re going from post training to runtime can network we and B be exported
runtime can network we and B be exported
runtime can network we and B be exported in order to use a better processing for
in order to use a better processing for
in order to use a better processing for example what is second question cannon
example what is second question cannon
example what is second question cannon can the network we can be be exported in
can the network we can be be exported in
can the network we can be be exported in order to reuse in embedded processing
order to reuse in embedded processing
order to reuse in embedded processing for example by exporting okay yeah
for example by exporting okay yeah
for example by exporting okay yeah before the first question which is the
before the first question which is the
before the first question which is the the online training I think it’s on the
the online training I think it’s on the
the online training I think it’s on the world map or maybe your do list of the
world map or maybe your do list of the
world map or maybe your do list of the nucleus but it’s currently it’s not
nucleus but it’s currently it’s not
nucleus but it’s currently it’s not supported but it’s possible that we’ll
supported but it’s possible that we’ll
supported but it’s possible that we’ll be supporting the online training where
be supporting the online training where
be supporting the online training where your neural networks will be gradually
your neural networks will be gradually
your neural networks will be gradually joined by the online data and second
joined by the online data and second
joined by the online data and second question is exporting yes you can export
question is exporting yes you can export
question is exporting yes you can export the trained parameter sets so that you
the trained parameter sets so that you
the trained parameter sets so that you can use the parameter assists to your
can use the parameter assists to your
can use the parameter assists to your une use cases such as the importing the
une use cases such as the importing the
une use cases such as the importing the parameters into the IOT devices or maybe
parameters into the IOT devices or maybe
parameters into the IOT devices or maybe you can even copy that data sets into
you can even copy that data sets into
you can even copy that data sets into different cloud record AWS to learn your
different cloud record AWS to learn your
different cloud record AWS to learn your predictions on database ok those
predictions on database ok those
predictions on database ok those questions are cut do anyone else has any
questions are cut do anyone else has any
questions are cut do anyone else has any questions no and thank you very much
questions no and thank you very much
questions no and thank you very much thank you so much
Be First to Comment