Formulating Discrete Probability Flow Through Optimal Transport

Authors: Pengze Zhang, Hubery Yin, Chen Li, Xiaohua Xie

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on the synthetic toy dataset and the CIFAR-10 dataset have validated the effectiveness of our proposed discrete probability flow.
Researcher Affiliation Collaboration Pengze Zhang Sun Yat-sen University zhangpz3@mail2.edu.cn Hubery Yin We Chat, Tencent Inc. hubery@tencent.com Chen Li We Chat, Tencent Inc. chaselli@tencent.com Xiaohua Xie Sun Yat-sen University xiexiaoh6@mail.edu.cn
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks, nor clearly labeled algorithm sections or code-like formatted procedures.
Open Source Code Yes Code is released at: https://github.com/Pangze Cheung/Discrete-Probability-Flow.
Open Datasets Yes Extensive experiments on the synthetic toy dataset and the CIFAR-10 dataset have validated the effectiveness of our proposed discrete probability flow.
Dataset Splits No The paper discusses sample generation and initial points but does not provide specific dataset split information (exact percentages, sample counts for train/validation/test, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiment.
Experiment Setup Yes Experiments are conducted on synthetic data using the same setup as SDDM [47], with the exception that we replaced the generator Q with Equation (26). ... Specifically, we set s = 0 and t = T, and sample 4,000 xts with 10 xss for each xt. ... We use the Euler s method to generate samples. Given the time step length ϵ, the transition probabilities for dimension l is: