Policy Gradient Methods - REINFORCE and REINFORCE with ACTOR-CRITIC

This repository contains an implementation of REINFORCE and ACTOR-CRITIC algorithms. The code is present in the src/ directory.

Requirements

• tensorflow < 2.0.0
• keras
• matplotlib
• gym
• numpy

Usage

In order to run the code, use the configuration file like lunarlander.py. Then run the following command to start training and evaluating the model.

python3 reinforce.py --env lunarlander.py  # REINFORCE
python3 a2c.py --env lunarlander_a2c.py  # A2C

Hope you find it helpful.