Our terms of service are changing. Learn more.

Publication

Latest

Short Introduction to Generative Adversarial Networks (GANs)

Last Updated on August 27, 2021 by Editorial Team

Author(s): Mugunthan

Machine Learning

General Adversarial Network (GAN) is a generative modeling approach using deep learning neural networks such as CNN.

There are two types of modeling techniques, i) Discriminative modeling and ii) generative modeling. Discriminative models are the typical ones that are used for classification in machine learning. They take input as features X (image, for image classification) and predict the output Y(probability of the image) for the given features. On the other hand, generative models output features X (image) given a random value.

Some of the generative models are Latent Dirichlet Allocation, or LDA, and the Gaussian Mixture Model, or GMM. Some Deep Learning generative models are Variational Autoencoders or VAE, General Adversarial Network, or GAN.

Simple representation of GAN’s work

How do GAN’s work?

GAN consists of two blocks, a generator, and a discriminator. The discriminator is a trained network that will be able to differentiate between real and fake images, While the generator is not trained. Using the random value(s), the generator will generate an image which then is fed to the discriminator. Then discriminator will give how fake/real is the generated image. This helps the generator to fine-tune the subsequent image generation, finally producing indistinguishable fake images.

By its name, you can see the generator and discriminator oppose each other for the generation of the new image.

Here are Images created by a GAN created by NVIDIA

These images are faces that don’t belong to any person, but are generated by GAN !.

Hope this blog at least clarifies what GAN is?


Short Introduction to Generative Adversarial Networks (GANs) was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

Published via Towards AI

Feedback ↓