Reinforcement Learning Project

This repository contains python notebooks used to develop an ATARI game playing agent for OpenAI gym using Keras-RL library. We implement the basic Deep Q-Learning model and then tried an interesting Transfer Learning approach. Our best 100-Episode Average score are listed below:

• Freeway - 21
• DemonAttack - 182
• AirRaid - 2995
• Centipede - 4332

Reinforcement Learning Turtorial

If you are new to Reinforcement learning or OpenAI Gym check out the Tutorial Notebook Notebook.

This was a class project and developed under an aggressive deadline and the notebooks have not been updated after the project so the files and code is not well organized. All the notebooks should run on Google Colab.