/MIS_solver

Primary LanguagePython

Solving NP-Hard Problems on Graphs with Extended AlphaGo Zero

Python implementation of the above paper for MaximumIndependentSet.

Note that this is an old version. The latest results in the paper were generated by C++ implementation.

Getting Started

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Directory Structure

  • environ: graph environment (MDP)
  • gcn: graph convolutional netoworks model
  • gin: graph isomophism networks model
  • policy: policy gradient method (REINFORCE)
  • trainer: train with REINFORCE
  • mcts: Monte Carlo Tree Search
  • utils: utils

Reference on Implementation