TorchNMTF

A collection of Nonnegative Matrix and Tensor Factorizations models with further applications to Multi-modal Data Analysis and Blind Source Separation and Sensor Fusion Problems, implemented with Pytorch.

Requirements

• PyTorch == 1.0.0 and it's dependences
• skimage
• Pandas
• ipypb

Usage

    see examples.ipynb

Here is a result of the inverse transform of a factorized image

Refereces

1. Andrzej Cichocki Rafal Zdunek Anh Huy Phan Shun‐Ichi Amari, Nonnegative Matrix and Tensor Factorizations

2. https://perso.telecom-paristech.fr/essid/teach/NMF_tutorial_ICME-2014.pdf

3. http://pmelchior.net/blog/proximal-matrix-factorization-in-pytorch.html

TODO

• Basic NMF model with 2 solvers
• ALS, HALS solvers, and advanced objectives
• More Examples
• Nonegative Tensor Factorisation
• Tests
• Setup.py
• Nonrandom initialization.