/Policy-Gradient-Reinforcement-Learning-on-a-Grid-World

Deep Reinforcement Learning implementation of Policy Gradient on a simple Grid-World problem using PyTorch.

Primary LanguagePythonMIT LicenseMIT

Policy Gradient Reinforcement Learning on a Grid World

About the Project

Deep Reinforcement Learning implementation of Policy Gradient on a simple Grid-World problem using PyTorch.

Example

In main.py define the environment and the agent, as well as the hyperparameters of the policy gradient network and run python3 main.py. The script saves a plot of the average rewards during training and validation. (See figure below for an example)

example

Prerequisites

  • pytorch
  • numpy
  • opencv
  • matplotlib.