Regular-Decision-Processes

Table of contents

• Development Setup
• Requirements
  • Download the glove data
  • Run
• Github Pages
• Report
• Slides

Development Setup

Requirements

$ conda create --name rdp python=3.8.10
$ conda activate rdp
$ pip install -r requirements.txt
$ cd ./nonmarkov-envs
$ pip install .

Run

$ cd ./nonmarkov-env/nonmarkov-env
$ python __main__.py --env ENV_NAME --only-mcts ONLY-MCTS --episode EPISODE --max-iterations MAX-ITERATION --min-samples MIN-SAMPLE --mixed MIXED

where

• ENV_NAME is the name of the domain, you can choose between: CheatMAB, RotatingMAB, RotatingMaze
• ONLY-MCTS choose if you want to perform only mcts algorithm, default is False
• EPISODE length of the episode of the env, default is 5
• MAX-ITERATION maximum number of iterations of the S3M algorithm, default is 4500
• MIN-SAMPLE minimum number of samples of the S3M algorithm, default is 100
• MIXED choose if you want to perform mixed S3M algorithm and MCTS saving figure, default is False

Github Pages

Report

Slides