Semi-Supervised Learning with Generative Adversarial Networks

TL;DR

This paper introduces a semi-supervised learning approach using GANs where the discriminator predicts class labels, improving data efficiency and sample quality.

Contribution

It extends GANs to semi-supervised learning by incorporating class label prediction into the discriminator, enhancing both classification and sample generation.

Findings

Improved data-efficient classification performance.

Higher quality sample generation compared to standard GANs.

Effective semi-supervised learning with limited labeled data.

Abstract

We extend Generative Adversarial Networks (GANs) to the semi-supervised context by forcing the discriminator network to output class labels. We train a generative model G and a discriminator D on a dataset with inputs belonging to one of N classes. At training time, D is made to predict which of N+1 classes the input belongs to, where an extra class is added to correspond to the outputs of G. We show that this method can be used to create a more data-efficient classifier and that it allows for generating higher quality samples than a regular GAN.