Generative Adversarial Networks (GANs)

Generative adversarial networks

Generative Adversarial Networks (GANs)

Introduction

The GAN generator

Generator: take latent multivar normal as input layer

Output of generator is of same dimension as input iamges

Generative NN: need probability distribution on input layer

Generator typically deconvolutional

Downside of gan. only discriminates between real and fake.

The GAN discriminator

Discriminator: use normal CNN for deep