Wasserstein of Wasserstein Loss for Learning Generative Models

Authors: Yonatan Dukler, Wuchen Li, Alex Lin, Guido Montufar

ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we present experiments demonstrating the effects and utility of WWGAN. We use the CIFAR-10 and 64 × 64 cropped-Celeb A image datasets.
Researcher Affiliation Academia 1Department of Mathematics and 2Department of Statistics, University of California, Los Angeles, CA 90095. 3Max Planck Institute for Mathematics in the Sciences, 04103 Leipzig, Germany. Correspondence to: Guido Mont ufar <montufar@math.ucla.edu>, Alex Tong Lin <atlin@math.ucla.edu>, Wuchen Li <wcli@math.ucla.edu>, Yonatan Dukler <ydukler@math.ucla.edu>.
Pseudocode Yes Algorithm 1 WWGAN Gradient Penalty. and Algorithm 2 Wasserstein gradient norm grad f(X) W .
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes We use the CIFAR-10 and 64 × 64 cropped-Celeb A image datasets.
Dataset Splits No The paper mentions using CIFAR-10 and Celeb A datasets but does not provide specific details on training, validation, and test splits (e.g., percentages or sample counts).
Hardware Specification No The paper mentions average epoch wall-clock times but does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions the use of 'ADAM optimizer' and neural networks, but it does not provide specific ancillary software details with version numbers (e.g., library or solver names with version numbers like Python 3.8, PyTorch 1.9).
Experiment Setup Yes The dimensionality of the latent variable of the generator is set at 128. Batch normalization is not applied to the discriminator, in order to avoid dependencies when computing the gradient penalties. The model is then trained with the ADAM optimizer with fixed parameters (β1, β2) = (0.9, 0).