calypsoherrera / lrs-cov-nn Goto Github PK

This repository is the official implementation of the paper Low Rank plus Sparse Decomposition of Covariance Matrices using Neural Network Parametrization.

Clone this git repo and cd into it.

git clone https://github.com/CalypsoHerrera/lrs-cov-nn.git
cd lrs-cov-nn

Then create a virtual environment. Use equivalent command for non debian based systems.

sudo apt-get install python3-venv
python3 -m venv py3
source py3/bin/activate
pip3 install --no-cache-dir -e .
pip install --upgrade pip

When contributing to project, if adding dependencies, please update the requirements.txt file.

pip freeze > requirements.txt

Generation of a low rank and of a sparse matrix. It is possible to choose the size of the matrix as well as the rank of the low rank matrix L and the sparsity of the sparse matrix S. Then the algorithm is run over the sum of them, Sigma = L+S.

python3 lrs/run_synthetic.py

Running the algo over a five hundred S&P500 stocks returns covariance matrix and generates images.

python3 lrs/run_sp500.py

Running the algo over a reale sate returns covariance matrix and generates images.

python3 lrs/run_realestate.py

This code can be used in accordance with the LICENSE.

If you use this library for your publications, please cite our paper: Low Rank plus Sparse Decomposition of Covariance Matrices using Neural Network Parametrization

@article{Baes2021Lowrank,
author    = {Baes, Michel and Herrera, Calypso and Neufeld, Ariel and Ruyssen, Pierre},
title     = {Low-Rank plus Sparse Decomposition of Covariance Matrices using Neural Network Parametrization
},
journal   = {CoRR},
volume    = {abs/2104.13669},
year      = {2021},
url       = {https://arxiv.org/abs/1908.00461}}

Last Page Update: 14/06/2021