Provably Fast Convergence of Independent Natural Policy Gradient for Markov Potential Games

Code for ''Provably Fast Convergence of Independent Natural Policy Gradient for Markov Potential Games'' (NeurIPS 2023)

Prerequisites

Here we list our running environment:

• PyTorch == 1.10.0
• docplex == 2.18.200
• numpy == 1.19.2
• matplotlib == 3.4.3

Tasks

We implement our proposed algorithms in the following environments:

• General Markov Potential Games (oracle)
• Congestion Game (oracle)

Execution

To run the experiments, simply execute the following commands, where ''xxx'' corresponds to the name of files.

python xxx.py

Credits

The code in this repository is adapted from the following works

• https://arxiv.org/pdf/2106.01969
• https://arxiv.org/abs/2206.05357