Graphical Generative Adversarial Networks

Authors: Chongxuan LI, Max Welling, Jun Zhu, Bo Zhang

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We implement our model using the Tensor Flow [1] library. In all experiments, we optimize the JS-divergence. We use the widely adopted DCGAN architecture [33] in all experiments to fairly compare Graphical-GAN with existing methods. We evaluate GMGAN on the MNIST [20], SVHN [30], CIFAR10 [19] and Celeb A [23] datasets. We evaluate SSGAN on the Moving MNIST [38] and 3D chairs [3] datasets. Qualitatively, Graphical-GAN can infer the latent structures and generate structured samples without any regularization, which is required by existing models [4, 43, 6, 42]; Quantitatively, Graphical-GAN can outperform all baseline methods [7 9] in terms of inference accuracy, sample quality and reconstruction error consistently and substantially.
Researcher Affiliation Academia Chongxuan Li licx14@mails.tsinghua.edu.cn Max Welling M.Welling@uva.nl Jun Zhu dcszj@mail.tsinghua.edu.cn Bo Zhang dcszb@mail.tsinghua.edu.cn Department of Computer Science & Technology, Institute for Artificial Intelligence, BNRist Center, THBI Lab, State Key Lab for Intell. Tech. & Sys., Tsinghua University. Correspondence to: J. Zhu. University of Amsterdam, and the Canadian Institute for Advanced Research (CIFAR).
Pseudocode Yes Algorithm 1 Local algorithm for Graphical-GAN
Open Source Code Yes Our source code is available at https://github.com/zhenxuan00/graphical-gan.
Open Datasets Yes We evaluate GMGAN on the MNIST [20], SVHN [30], CIFAR10 [19] and Celeb A [23] datasets. We evaluate SSGAN on the Moving MNIST [38] and 3D chairs [3] datasets.
Dataset Splits No The paper mentions 'test samples' when discussing evaluation but does not provide specific details on the percentages or counts for training, validation, and test splits, nor does it cite a specific split methodology used for these standard datasets.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory specifications, or cloud instance types) used for running the experiments. It only mentions the use of the Tensor Flow library.
Software Dependencies No The paper mentions using the 'Tensor Flow [1] library' but does not provide a specific version number for TensorFlow or any other software dependencies.
Experiment Setup No The paper states that 'In all experiments, we optimize the JS-divergence. We use the widely adopted DCGAN architecture [33] in all experiments' but does not provide specific hyperparameters such as learning rates, batch sizes, number of epochs, or detailed optimizer settings in the main text.