arXiv (Cornell University) · 2016 · 1.4K citations · 0 references
ComputingComputer ScienceArtificial IntelligenceComputer VisionMachine LearningDeep LearningGenerative ModelingNeural NetworksRepresentation LearningSemi-supervised LearningGenerative Adversarial NetworksGenerative Image SynthesisSemi-supervised ClassificationSemi-supervised Image ClassificationSemi-supervised Representation Learning
The study departs from typical generative modeling by focusing on training GANs for tasks beyond likelihood maximization and unsupervised learning. The authors aim to develop new architectural features and training procedures for GANs to enhance semi‑supervised learning and realistic image generation. They implement these features and procedures within the GAN framework, applying them to semi‑supervised learning and image‑generation tasks. The resulting model achieves state‑of‑the‑art semi‑supervised classification on MNIST, CIFAR‑10, and SVHN, produces images that fool human judges in visual Turing tests, and generates high‑resolution ImageNet samples with recognizable class features.
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.