Large-scale optimal transport map estimation using projection pursuit

Authors: Cheng Meng, Yuan Ke, Jingyi Zhang, Mengrui Zhang, Wenxuan Zhong, Ping Ma

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, PPMM is computationally easy and converges fast. We assess its finite sample performance through the applications of Wasserstein distance estimation and generative models.
Researcher Affiliation Academia 1Department of Statistics, University of Georgia {cheng.meng25, yuan.ke, jingyi.zhang25, mengrui.zhang, wenxuan, pingma }@uga.edu
Pseudocode Yes Algorithm 1 Select the most informative projection direction using SAVE; Algorithm 2 Projection pursuit Monge map (PPMM)
Open Source Code No The paper does not provide a direct link or explicit statement about the availability of the source code for the PPMM method described in the paper. It only mentions the R package 'transport' used for comparison methods.
Open Datasets Yes We study the MNIST dataset, which contains 60,000 training images and 10,000 testing images of hand written digits. The Google Doodle dataset4 https://quickdraw.withgoogle.com/data
Dataset Splits Yes We randomly split the data into a training set and a validation set of equal sample sizes.
Hardware Specification Yes The experiments are implemented by an Intel 2.6 GHz processor.
Software Dependencies No The paper mentions 'R package transport [46]' but does not provide specific version numbers for this or any other critical software dependencies.
Experiment Setup Yes We set n = 10, 000, d = {10, 20, 50}, µX = 2, µY = 2, ΣX = 0.8|i j|, and ΣY = 0.5|i j|, for i, j = 1, . . . , d. For all three methods, we set the maximum number of iterations to be 200. The dimension of the latent space is set to be 16. we set the termination criteria to be a hard threshold, i.e., 10 5.