This is the code for the paper "Provable Alternating Gradient Descent for Non-negative Matrix Factorization with Strong Correlations".

1. Core code

The code consists of the following.

competiors.py is the code for the competior algorithms.
AND.py is the code for our algorithm AND (Alternative Non-negative gradient
Descent).

data_io.py is the code for loading/generating data.
compute_error.py is the code for computing the error.


2. Running the code

demo.sh is a simple example on how to use the core code.