Linear optimal partial transport embedding

Authors: Yikun Bai, Ivan Vladimir Medri, Rocio Diaz Martin, Rana Shahroz, Soheil Kolouri

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

Reproducibility Variable Result LLM Response
Research Type Experimental We test the approximation performance of OPT using LOPT. Given K empirical measures {µi}K i=1, for each pair (µi, µj), we compute OPTλ(µi, µj) and LOPTµ0,λ(µi, µj) and the mean or median of all pairs (µi, µj) of relative error defined as |OPTλ(µi, µj) LOPTµ0,λ(µi, µj)| / OPTλ(µi, µj) . For our experiments, we created K point sets of size N = 500 for K different Gaussian distributions in R2.
Researcher Affiliation Academia 1Department of Computer Science, Vanderbilt University, 366 Featheringill, Nashville, TN 37240, USA 2Department of Mathematics, Vanderbilt University, 1326 Stevenson Center, Nashville, TN 37240, USA.
Pseudocode No The paper provides mathematical formulations and descriptions of methods but does not include structured pseudocode or algorithm blocks.
Open Source Code Yes Our code is available at https://github.com/ Baio0/Linear OPT.
Open Datasets Yes Point Cloud Interpolation: We test OT geodesic, LOT geodesic, HK geodesic, LHK geodesic, OPT interpolation, and LOPT interpolation on the point cloud MNIST dataset. ... PCA analysis: We compare the results of performing PCA on the embedding space of LOT, LHK and LOPT for point cloud MNIST. We take 900 digits from the dataset corresponding to digits 0, 1 and 3 in equal proportions.
Dataset Splits No The paper mentions using the MNIST dataset but does not specify the exact training, validation, and test splits (e.g., percentages or sample counts) used for the experiments.
Hardware Specification Yes The experiment was conducted on a Linux computer with AMD EPYC 7702P CPU with 64 cores and 256GB DDR4 RAM. ... The experiments are conducted on a Linux computer with AMD EPYC 7702P CPU with 64 cores and 256GB DDR4 RAM.
Software Dependencies No The paper mentions "Python OT" as the LP solver used but does not specify its version number or any other software dependencies with version details.
Experiment Setup Yes For our experiments, we created K point sets of size N = 500 for K different Gaussian distributions in R2. In particular, µi N(mi, I), where mi is randomly selected such that mi = 3 for i = 1, ..., K. For the reference, we picked an N point representation of µ0 N(m, I) with m = P mi/K. ... In OPT and LOPT interpolation, we set λ = 20; in HK and LHK, we set the scaling to be 2.5. ... We test for η = 0, 0.5, 0.75 (see Figure 8 in the Appendix H).