Decentralized Projected Riemannian Gradient Descent (DPRGD)

This is a repository for the paper Decentralized Online Riemannian Optimization with Dynamic Environments. Our experiments cover two Hadamard manifolds: hyperbolic spaces and the space of symmetric prositive definite (SPD) matrices.

Introduction to the files

1. hyperboloid.py and spd.py provide basic functions for hyperbolic spaces and the space of SPD matrices.
2. DPRGD.ipynb collects all the experiments.