Name: 沒有博士學位的TensorFlow和深度學習，第一部分（Google Cloud Next '17）。 (TensorFlow and Deep Learning without a PhD, Part 1 (Google Cloud Next '17))
Uploaded: 2021-01-14T06:50:37.000Z
Duration: 55 min 52 s
Description: 【看影片學英語】數萬部 YouTube 影片，搭配英漢字典即點即查，輕鬆掌握單字發音與用法，長久累積看電影不必再看字幕。

So thank you for coming in such great numbers

Apologies, it's quite late in the afternoon.

I will need all your brains for this session because today,

I want with you to build a neural network.

So no, I don't need your brains to build on, no brain

But it's a crash course to get developers up

to speed on machine learning and deep learning and neural

The dataset we will be using is a very classical one.

Academia has been working on this dataset for the past 20

So you should go to the website where it's hosted.

You will actually see 20 years of research papers

and that's what we will do together today.

We'll go on this dataset trying to build a network that

recognizes this hand-written digits from the simplest

possible network all the way to 99% accuracy.

Who has done some work with neural networks before?

So feel free to help me and I hope this will not

will at least be a good introduction to TensorFlow.

But if you have never done anything with neural networks,

that's fine and I will explain everything from the start.

So this is the simplest possible neural network

we can imagine to recognize our hand-written digits.

So the digits, they come as 28 by 28 pixel images

and the first thing we do is that we flatten

all those pixels into one big vector of pixels

What a neuron does is always the same thing.

A neuron does a weighted sum of all of its inputs,

It adds another constant that is called a bias.

That's just an additional degree of freedom.

And then it will feed this sum through an activation function.

And that is just a function-- number in, transform,

We will see several of those activation functions

and the one thing they have in common in neural networks

Well, simply because we are classifying those digits

We are trying to recognize a zero, a one, a two,

So what we are hoping for here is that one of those neurons

will light up and tell us, with a very strong output,

And for that, since this is a classification problem,

we are going to use a very specific activation

tell us works really well on classification problems.

So what you do is that you make all those weighted sums,

then you elevate that to the exponential.

and divide it by its norm so that you get

And those values, you will be able to interpret them

as probabilities, probabilities of this being an eight, a one,

which is the Euclidean normal would work just as well.

You see, it's an exponential so it's a very steeply increasing

It will pull the data apart, increase the differences,

and when you divide all of that, when you normalize

the whole vector, you usually end up with one of the values

being very close to one and all the other values

So it's a way of pulling the winner out on top

without actually destroying the information.

So now we need to formalize this using a matrix multiply.

I will remind you of what a matrix multiply is,

are going to do this for a batch of 100 images at a time.

字幕列表影片播放

沒有博士學位的TensorFlow和深度學習，第一部分（Google Cloud Next '17）。 (TensorFlow and Deep Learning without a PhD, Part 1 (Google Cloud Next '17))

recognize

trick

basically

bias