Unsupervised Ground Metric Learning Using Wasserstein Singular Vectors

Authors: Geert-Jan Huizing, Laura Cantini, Gabriel Peyré

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we showcase Wasserstein Singular Vectors on a single-cell RNA-sequencing dataset. Section 5 demonstrates the potential of Wasserstein Singular Vectors compared to ad-hoc applications of Optimal Transport, by studying a single-cell RNA-sequencing dataset.
Researcher Affiliation Academia 1D epartement de math ematiques et applications de l Ecole Normale Sup erieure, CNRS, Ecole Normale Sup erieure, Universit e PSL, 75005, Paris, France 2Computational Systems Biology Team, Institut de Biologie de l Ecole Normale Sup erieure, CNRS, INSERM, Ecole Normale Sup erieure, Universit e PSL, 75005, Paris, France.
Pseudocode No The paper describes algorithms in prose and mathematical formulations but does not include explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes A Python package implementing all algorithms in this paper is available at github.com/gjhuizing/wsingular.
Open Datasets Yes A commonly analyzed sc RNA-seq dataset is the PBMC 3k dataset produced by 10X Genomics, obtained through the function pbmc3k of Scanpy (Wolf et al., 2018).
Dataset Splits No The paper describes the dataset and its preprocessing but does not explicitly state train/validation/test splits or cross-validation methodology for the experiments.
Hardware Specification Yes CPU computations were performed on a Dell Latitude 5420 with an 8 core 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz CPU. GPU computations were performed on Nvidia V100 SXM2 32 Go GPUs.
Software Dependencies No The paper mentions software like 'Python', 'POT library', and 'Scikit-learn' but does not provide specific version numbers for these dependencies.
Experiment Setup Yes Wasserstein Singular Vectors (τ = 0.001, ε = 0.1), approached by 15 power iterations