stair-lab / dynamicCov

A Nonconvex Framework for Structured Dynamic Covariance Recovary

Table of contents

• General info
• Requirements
• Setup
• Examples
• License
• References

Genral Info

This repo is the implementation of the paper [1] along with the competing methods listed in Table 1 in [1]. Please check Table 1 for details.

Requirements

Install the ssm package to run comppeting methods. It is recommended to install Anaconda and create a new environment (python>=3.7) to run this code.

conda create --name snscov python=3.7
conda activate snscov

Setup

python setup.py install

Examples

Some simulation examples.

1. Run Simulation 1. Variations of waveforms

cd examples/simulation1
python simulation1.py 

Models are stored in folder ./results and plots are stored in ./figures. Specify the waveform by using --waveform

2. Run Simulation 2. Comparison with other mehtods

cd examples/comparison1
python compare_all.py

Test different number of test subjects by running the follow bash file. Select the waveform (sine, square, mixing) by changing --waveform in compare.sh

cd examples/comparison1
bash compare.sh

3. Run Simulation 3. Comparison with other mehtods (high-dimensional data)

cd examples/comparison2
python test_large_scale.py

License

Distributed under the MIT License. See LICENSE.txt for more information.

References

[1] Tsai, K., Kolar, M., & Koyejo, O. (2022). A Nonconvex Framework for Structured Dynamic Covariance Recovery. Journal of Machine Learning Research, 23(200), 1-91.