wtf: Wasserstein Tensor Factorisation

A unified framework for non-negative matrix and tensor factorisations with a smoothed Wasserstein loss

This repository contains a basic implementation of the method described in the article "A unified framework for non-negative matrix and tensor factorisations with a smoothed Wasserstein loss".

Requirements

• PyTorch
• Tensorly
• PythonOT
• CUDA-compatible GPU (e.g. use Colab) for efficient autodiff

Instructions

• Clone this repo: git clone https://github.com/zsteve/wtf.git
• Import the wtf module using

  import sys
  sys.path.insert(0, "/content/wtf/src")
  import wtf

• ???
• Profit

Example

From the paper

A notebook for Figures 3, 4, 5 is located in the examples/ directory.

Amino acids fluorescence dataset

A notebook for the amino acids fluorescence dataset of Andersson and Bro is available as examples/amino.ipynb.

Thanks to Shoaib Bin Masud (Tufts) and Anna Konstorum (Yale) for bringing this dataset to my attention.

Citing

• If you find this work relevant to your research project, please cite the preprint

Zhang, S. A unified framework for non-negative matrix and tensor factorisations with a smoothed Wasserstein loss, arXiv preprint arXiv:2104.01708, 2021