Improved Techniques for Training GANs

Authors: Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, Xi Chen, Xi Chen

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 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 Image Net samples with unprecedented resolution and show that our methods enable the model to learn recognizable features of Image Net classes.
Researcher Affiliation Industry Tim Salimans tim@openai.com Ian Goodfellow ian@openai.com Wojciech Zaremba woj@openai.com Vicki Cheung vicki@openai.com Alec Radford alec@openai.com Xi Chen peter@openai.com
Pseudocode No The paper describes methods and processes verbally and with diagrams, but it does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes All code and hyperparameters may be found at https://github.com/openai/improved-gan.
Open Datasets Yes We performed semi-supervised experiments on MNIST, CIFAR-10 and SVHN, and sample generation experiments on MNIST, CIFAR-10, SVHN and Image Net.
Dataset Splits Yes The MNIST dataset contains 60, 000 labeled images of digits. We perform semi-supervised training with a small randomly picked fraction of these, considering setups with 20, 50, 100, and 200 labeled examples. Results are averaged over 10 random subsets of labeled data, each chosen to have a balanced number of examples from each class. The remaining training images are provided without labels.
Hardware Specification No We extensively modified a publicly available implementation of DCGANs2 using Tensor Flow [28] to achieve high performance, using a multi-GPU implementation. However, no specific GPU model or other hardware specs are given.
Software Dependencies No We extensively modified a publicly available implementation of DCGANs2 using Tensor Flow [28] to achieve high performance, using a multi-GPU implementation. No version numbers for TensorFlow or any other specific library/solver are provided.
Experiment Setup Yes Our networks have 5 hidden layers each. We use weight normalization [21] and add Gaussian noise to the output of each layer of the discriminator. (MNIST) and For the discriminator in our GAN we use a 9 layer deep convolutional network with dropout and weight normalization. The generator is a 4 layer deep CNN with batch normalization. (CIFAR-10)