Deep Reinforcement Learning

Q Learning and Deep Q-Learning to train agents that plays different games

I implemented Q-table and Deep Q-Learning algorithms to play various games, including Cart Pole, Flappy Bird, and Breakout, using the OpenAI Gym. This project provided hands-on experience in reinforcement learning and sequence models. The input to these algorithms was either raw pixel values or game-specific state information, enabling the models to learn and play each game.

Cart Pole Screenshot

Mountain Car Screenshot

Custom environment screenshot

Key Features:

Implemented Q-table-based reinforcement learning for small environments.
Developed Deep Q-Learning and Double Q-Learning algorithms.
Created a custom environment and game.
Trained models on an Atari game from the OpenAI Gym.

Tools and Resources:

Python
PyTorch
Anaconda Python (for local work)
VT ARC cluster (for GPU access)

Source Code

The complete source code for this project is available on GitHub.

Cover Image Credit: Cool Math Games Blog