config_corr

Configuration model for correlation/covariance matrices

This package contains two algorithms.

• Algorithm DMCC solves a convex optimization problem and usually better than the other naive gradient descent algorithm. Therefore, we recommend DMCC over the naive gradient descent algorithm.
• DMCC, including the performance test, is documented in the accompanying DMCC-algorithm.pdf.
• DMCC is provided in Python only.
• The naive gradient descent algorithm is provided in Python and MATLAB.

Install using pip

pip install configcorr

Install from source

Package configcorr has the following dependencies:

Python >= 3.4
setuptools >= 1.4
NumPy >= 1.8
CVXPy >= 1.0

These packages need to be installed manually.

Clone the configcorr git repository by running

git clone https://github.com/naokimas/config_corr.git

Then, navigate to the top-level of the cloned directory and run

python setup.py install

DMCC algorithm in Python

python3 test_dmcc.py

• test_dmcc.py calls max_ent_config_dmcc.py

Naive gradient descent algorithm in Python

python3 test_naive_gradient_descent.py

• test_naive_gradient_descent.py calls max_ent_config_naive_gradient_descent.py

Naive gradient descent algorithm in MATLAB

test_naive_gradient_descent

• test_naive_gradient_descent.m calls max_ent_config_naive_gradient_descent.m
• The data set to be used should be specified within test_naive_gradient_descent.m
• Variable curr_dir should be modified according to where you place the data set.