iamalexkorotin/Wasserstein2Barycenters

PyTorch implementation of the paper "Continuous Wasserstein-2 Barycenter Estimation without Minimax Optimization" (ICLR 2021)

29
/ 100
Experimental

This project helps researchers and practitioners in machine learning and data science who need to find an 'average' or 'central' distribution from multiple input distributions, especially when dealing with high-dimensional data like image color palettes. It takes several continuous probability distributions as input and efficiently calculates their Wasserstein-2 barycenter, which represents a geometrically meaningful average. This is particularly useful for scientists working with complex data averaging and transformation tasks.

No commits in the last 6 months.

Use this if you need to compute a continuous Wasserstein-2 barycenter for high-dimensional distributions, especially for tasks like averaging color palettes or other complex data distributions, and prefer an efficient, non-minimax approach.

Not ideal if your task involves discrete distributions, or if you are not familiar with deep learning frameworks like PyTorch, as the implementation is GPU-based and uses input convex neural networks.

Machine Learning Research Data Science Image Processing Distribution Averaging Optimal Transport
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 7 / 25
Maturity 16 / 25
Community 6 / 25

How are scores calculated?

Stars

34

Forks

2

Language

Jupyter Notebook

License

MIT

Last pushed

Jun 17, 2022

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/iamalexkorotin/Wasserstein2Barycenters"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.