A deep reinforcement learning implementation that masters the classic Atari game Pong using PyTorch and Deep Q-Learning. 🎯 Project Overview This project implements a Deep Q-Learning Neural Network (DQN) that learns to play Atari's Pong from scratch. The agent learns by interacting with the game environment, developing strategies to defeat the built-in AI opponent. 🚀 Features
Deep Q-Learning implementation using PyTorch Frame preprocessing for efficient learning Experience replay buffer for improved training stability Epsilon-greedy exploration strategy Performance metrics tracking and visualization Trained model checkpoints
🛠️ Prerequisites
-Python 3.8+
-PyTorch
-OpenAI Gym[atari]
-NumPy
-Matplotlib
This project is part of a Reinforcement Learning course I took during my Master's in Artificial Intelligence at IU. Here, you will find the code and files for a reinforcement learning agent playing pong atari.
This project is self-funded. I was not able to complete the training due to the cost of computing resources. If anyone has further suggestions or further feedback, please send me an email at kivroglouvivi@gmail.com
If you have a problem installing the OpenCV library in a Jupyter Notebook, try: !pip install opencv-python-headless