Name: 神經結構化學習--第四部分：影像分類的對抗性學習 (Neural Structured Learning - Part 4: Adversarial learning for image classification)
Uploaded: 2021-01-14T10:01:04.000Z
Duration: 8 min 6 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。

DA-CHENG JUAN: Welcome to the fourth episode of Neural

about learning with implicit structured signals constructed

I'm Da-Cheng, and I'm going to be your guide.

You do not need to know a lot about adversarial learning

And we will learn the concept along the way.

This is the example we mentioned in the first episode,

In reality, there are usually other similar images

a structure that represents the similarity among all

And a neural unstructured learning framework jointly

optimizes the sample features and the structured signals

existing among the samples to learn a better neural net.

You may want to ask, what if there's no explicit structure

One approach is to construct the structure dynamically

An adversarial neighbor is a modified version

at misleading the neural net to make it output

is, how to generate such adversarial neighbors?

We craft a small amount of carefully designed

perturbation, usually based on the reverse gradient direction,

and apply that perturbation to the original sample.

Usually, human eyes cannot tell the difference between the two.

by the adversarial neighbor and classifies it incorrectly

This is because the small perturbation applied

on the sample successfully confuses the model,

After the adversarial neighbor is generated,

with its adversarial neighbor to dynamically construct

Then this structure can be used in the neural structure

why do we want to have a structure connecting the sample

In a neural structure learning framework,

the neural net learns to maintain a structure

by keeping the similarity between a sample

So essentially, this is telling the neural net

So please keep their similarity, and don't be confused

there are TensorFlow library and functions

that you can use to generate adversarial neighbors.

can use to enable easy-to-use end-to-end training

of this library and APIs, please visit our website.

to see how the adversarial learning works.

Say we want to train a neural net to recognize

In the next slide, we are going to write some Python

code to design this neural net and train it

the handwritten digits using adversarial learning.

that contains the image of the handwritten digits

so that they will stay in the range from 0 to 1.

Next, we build a neural net by using Keras APIs.

You can use Keras sequential APIs, functional APIs,

Here we invoke the Keras APIs from the neural structure

learning framework to enable adversarial learning.