Exploiting Chain Rule and Bayes' Theorem to Compare Probability Distributions

Authors: Huangjie Zheng, Mingyuan Zhou

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental On a wide variety of benchmark datasets for generative modeling, substituting the default statistical distance of an existing generative adversarial network with CT is shown to consistently improve the performance. Py Torch code is provided.
Researcher Affiliation Academia Huangjie Zheng Department of Statistics & Data Science The University of Texas at Austin Austin, TX 78712 huangjie.zheng@utexas.edu Mingyuan Zhou Mc Combs School of Business The University of Texas at Austin Austin, TX 78712 mingyuan.zhou@mccombs.utexas.edu
Pseudocode No The paper presents mathematical derivations and equations but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes Our code is available at https://github.com/Jeg Zheng/CT-pytorch.
Open Datasets Yes To demonstrate the use of CT, we apply it to train implicit (or explicit) distributions to model both 1D and 2D toy data, MNIST digits, and natural images. ...We consider three widely-used image datasets, including CIFAR-10 [44], Celeb A [45], and LSUN-bedroom [46] for general evaluation, as well as Celeb A-HQ [47], FFHQ [48] for evaluation in high-resolution.
Dataset Splits No The paper does not explicitly provide specific training, validation, and test dataset splits (e.g., percentages or exact counts) for reproducibility.
Hardware Specification No The Acknowledgments section mentions 'the Texas Advanced Computing Center (TACC) for providing HPC resources', but no specific hardware details (e.g., GPU models, CPU types) are provided for the experiments.
Software Dependencies No The paper mentions 'Py Torch code is provided' but does not specify its version or any other software dependencies with version numbers.
Experiment Setup Yes On the choice of ρ for natural images: In previous experiments, we fix ρ = 0.5 by default... We set X with 5000 samples and the mini-batch size as N = 100.