generative adversarial networks - deep learning
TRANSCRIPT
![Page 1: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/1.jpg)
GENERATIVE ADVERSARIAL NETWORKS
11785- Introduction to Deep Learning
AKSHAT GUPTAKUSHAL SAHARAN
![Page 2: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/2.jpg)
–Yann LeCun
“This (GANS), and the variations that are now being proposed is the most interesting idea in the last 10
years in ML, in my opinion”
![Page 4: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/4.jpg)
WHAT ARE GANS?
Generative Adversarial Networks
![Page 5: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/5.jpg)
WHAT ARE GANS?
Generative Adversarial Networks
Generative ModelsWe try to learn the underlying the distribution
from which our dataset comes from.Eg: Variational AutoEncoders (VAE)
![Page 6: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/6.jpg)
WHAT ARE GANS?
Generative Adversarial Networks
Generative ModelsWe try to learn the underlying the distribution
from which our dataset comes from.Eg: Variational AutoEncoders (VAE)
Encoderx ~ (training set) P(z) Decoder
![Page 7: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/7.jpg)
WHAT ARE GANS?
Generative Adversarial Networks
Generative ModelsWe try to learn the underlying the distribution
from which our dataset comes from.Eg: Variational AutoEncoders (VAE)
Adversarial TrainingGANS are made up of two competing networks (adversaries)
that are trying beat each other.
![Page 8: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/8.jpg)
WHAT ARE GANS?
Generative Adversarial Networks
Generative ModelsWe try to learn the underlying the distribution
from which our dataset comes from.Eg: Variational AutoEncoders (VAE)
Adversarial TrainingGANS are made up of two competing networks (adversaries)
that are trying beat each other.
Neural Networks
![Page 9: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/9.jpg)
WHAT ARE GANS?
P(z) Generator Generated Data
![Page 10: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/10.jpg)
WHAT ARE GANS?
P(z) Generator Generated Data
Discriminator Real/Fake?
![Page 11: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/11.jpg)
WHAT ARE GANS?
P(z) Generator Generated Data
Discriminator Real/Fake?
RealData
![Page 12: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/12.jpg)
HOW TO TRAIN A GAN?
![Page 13: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/13.jpg)
HOW TO TRAIN A GAN?At t = 0,
GeneratorLatent Vector
GeneratedImage (fake image)
Generated Data
Discriminator Real/Fake?
GivenTraining
Data
(fake data)
(Real data)
![Page 14: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/14.jpg)
HOW TO TRAIN A GAN?At t = 0,
GeneratorLatent Vector
GeneratedImage (fake image)
Generated Data
Discriminator Real/Fake?
GivenTraining
Data
(fake data)
(Real data)
BinaryClassifier
![Page 15: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/15.jpg)
HOW TO TRAIN A GAN?Which network should I train first?
![Page 16: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/16.jpg)
HOW TO TRAIN A GAN?Which network should I train first?
Discriminator!
![Page 17: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/17.jpg)
HOW TO TRAIN A GAN?
Which network should I train first?Discriminator!
But with what training data?
![Page 18: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/18.jpg)
HOW TO TRAIN A GAN?
Which network should I train first?Discriminator!
But with what training data?The Discriminator is a Binary classifier.The Discriminator has two class - Real and Fake.The data for Real class if already given: THE TRAINING DATAThe data for Fake class? -> generate from the Generator
![Page 19: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/19.jpg)
HOW TO TRAIN A GAN?What’s next? -> Train the Generator
But how? What’s our training objective?
![Page 20: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/20.jpg)
HOW TO TRAIN A GAN?What’s next? -> Train the Generator
But how? What’s our training objective?Generate images from the Generator
such that they are classified incorrectly by the Discriminator!
![Page 21: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/21.jpg)
HOW TO TRAIN A GAN?
Discriminator
Step 1:Train the Discriminator using the current ability
of the Generator.
![Page 22: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/22.jpg)
HOW TO TRAIN A GAN?
GeneratorDiscriminator
Step 1:Train the Discriminator using the current ability
of the Generator.
Step 2:Train the Generator
to beatthe Discriminator.
![Page 23: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/23.jpg)
HOW TO TRAIN A GAN?
GeneratorDiscriminator
Step 1:Train the Discriminator using the current ability
of the Generator.
Step 2:Train the Generator
to beatthe Discriminator.
Generate images from the Generator such that they are classified incorrectly by the Discriminator!
![Page 24: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/24.jpg)
HOW TO TRAIN A GAN?
GeneratorDiscriminator
Step 1:Train the Discriminator using the current ability
of the Generator.
Step 2:Train the Generator
to beatthe Discriminator.
![Page 25: GENERATIVE ADVERSARIAL NETWORKS - Deep Learning](https://reader030.vdocuments.site/reader030/viewer/2022012503/617dc4755a0dc13d6d73919d/html5/thumbnails/25.jpg)
MNIST AND FASHION-MNIST