Improved Techniques for Training GANs

TL;DR

This paper introduces new architectural features and training procedures for GANs, achieving state-of-the-art semi-supervised classification results and generating highly realistic images, including high-resolution ImageNet samples.

Contribution

It presents novel techniques that improve GAN training, leading to better semi-supervised learning performance and more realistic image generation, including high-resolution ImageNet outputs.

Findings

State-of-the-art semi-supervised classification on MNIST, CIFAR-10, SVHN

Generated images are indistinguishable from real data in visual Turing tests

Enables high-resolution ImageNet image generation with recognizable features

Abstract

We present a variety of new architectural features and training procedures that we apply to the generative adversarial networks (GANs) framework. We focus on two applications of GANs: semi-supervised learning, and the generation of images that humans find visually realistic. Unlike most work on generative models, our primary goal is not to train a model that assigns high likelihood to test data, nor do we require the model to be able to learn well without using any labels. Using our new techniques, we achieve state-of-the-art results in semi-supervised classification on MNIST, CIFAR-10 and SVHN. The generated images are of high quality as confirmed by a visual Turing test: our model generates MNIST samples that humans cannot distinguish from real data, and CIFAR-10 samples that yield a human error rate of 21.3%. We also present ImageNet samples with unprecedented resolution and show that our methods enable the model to learn recognizable features of ImageNet classes.