POT: Python Optimal Transport

Authors: Rémi Flamary, Nicolas Courty, Alexandre Gramfort, Mokhtar Z. Alaya, Aurélie Boisbunon, Stanislas Chambon, Laetitia Chapel, Adrien Corenflos, Kilian Fatras, Nemo Fournier, Léo Gautheron, Nathalie T.H. Gayraud, Hicham Janati, Alain Rakotomamonjy, Ievgen Redko, Antoine Rolet, Antony Schutz, Vivien Seguy, Danica J. Sutherland, Romain Tavenard, Alexander Tong, Titouan Vayer

JMLR 2021 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The POT library takes advantage of Python to make Optimal Transport accessible to the machine learning community. It provides state-of-the-art algorithms to solve the regular OT optimization problems, and related problems such as entropic Wasserstein distance with Sinkhorn algorithm or barycenter computations. ... Solvers are illustrated on one or two dimensional simulated data, as well as some real world problems such as color transfer.
Researcher Affiliation Collaboration R emi Flamary EMAIL Nicolas Courty EMAIL Alexandre Gramfort EMAIL Mokhtar Z. Alaya EMAIL Aur elie Boisbunon EMAIL Stanislas Chambon EMAIL Laetitia Chapel EMAIL Adrien Corenflos EMAIL Kilian Fatras EMAIL Nemo Fournier EMAIL L eo Gautheron EMAIL Nathalie T.H. Gayraud EMAIL Hicham Janati EMAIL Alain Rakotomamonjy EMAIL Ievgen Redko EMAIL Antoine Rolet EMAIL Antony Schutz EMAIL Vivien Seguy EMAIL Danica J. Sutherland EMAIL Romain Tavenard EMAIL Alexander Tong EMAIL Titouan Vayer EMAIL
Pseudocode No The paper includes 'Sample 1: OT matrix POT syntax' and 'Sample 2: OT barycenters POT syntax', which are Python code snippets demonstrating library usage, not structured pseudocode or algorithm blocks describing the underlying methods.
Open Source Code Yes This toolbox, named POT for Python Optimal Transport, is open source with an MIT license. ... It is hosted on Git Hub1 and a public mailing list is available to the community to work on or follow the toolbox. ... 1. https://github.com/Python OT/POT
Open Datasets No Solvers are illustrated on one or two dimensional simulated data, as well as some real world problems such as color transfer. The paper mentions simulated and real-world data but does not provide specific names, citations, or access information for publicly available datasets.
Dataset Splits No The paper uses simulated data and mentions real-world problems like color transfer, but it does not provide specific details on dataset splits (e.g., train/test/validation percentages or counts) for any experiments.
Hardware Specification No The paper mentions 'GPU acceleration with a Cu Py implementation' and that 'Geom Loss [...] is a more specific toolbox for solving very large scale Sinkhorn on CPU and GPU', but does not provide specific hardware details (like model numbers or processors) used for the experiments described in this paper.
Software Dependencies Yes POT relies only on open source libraries such as Num Py (Harris et al., 2020) and Sci Py (Virtanen et al., 2020) for linear algebra and optimization problems. Visualization, mostly used in the examples, requires the Matplotlib (Hunter, 2007) visualization library. The Cython (Behnel et al., 2011) framework was used to provide a simple Python wrapper around the C++ code.
Experiment Setup No The paper describes the functionalities and usage of the POT library with code examples, but does not provide specific experimental setup details such as hyperparameter values, training configurations, or system-level settings for experiments.