Switch to Human Mode

About This Project

This is an implementation of Deep Q-Network (DQN) playing Breakout. DQN is a groundbreaking reinforcement learning algorithm introduced by DeepMind in their seminal paper "Human-level control through deep reinforcement learning" (Nature, 2015). The algorithm combines Q-learning with deep neural networks to learn optimal gameplay directly from pixels.

In this implementation:

• The model was trained offline using Python and TensorFlow - you can view the training code here
• The trained model is loaded and run in this web app using TensorFlow.js with WebGPU acceleration for inference
• The agent learns to position the paddle to bounce the ball and break bricks by maximizing expected future rewards

Try switching between AI and human mode to compare your skills against the trained agent!