This repo contains the code for the paper Decentralized Online Riemannian Optimization with Dynamic Environments.

• Our simulation studies cover hyperbolic spaces and the space of symmetric prositive definite (SPD) matrices.
• Our data applications involves environmental monitoring using the FLUXNET2015 dataset.

Create and activate conda environments and install necessary dependencies.

conda create --name opt python=3.10
conda activate opt
pip install -r requirements.txt

Download FLUXNET2015 dataset into the folder data/zip.

Run the following code to unzip, select, and save raw data into data/raw.

python src/data/data_loader.py

Next, we compute weekly correlation/covariance matrices and save the processed data into data/processed.

python src/data/data_processor.py

• simulation.ipynb contains the simulation experiments
• data.ipynb contains real data analysis using the FLUXNET2015 dataset

@article{chen2024decentralized,
  title={Decentralized Online Riemannian Optimization with Dynamic Environments},
  author={Chen, Hengchao and Sun, Qiang},
  journal={arXiv preprint arXiv:2410.05128},
  year={2024}
}