About
This course explores the core principles of AlphaZero, DeepMind's groundbreaking AI, and applies them to simpler board games. Gain hands-on experience building AI models from scratch using GPUs and cloud environments.
Content
- Implement the Tic Tac Toe game logic to enable agents to learn how to play.
- Implement the MCTS algorithm to estimate the best moves for game players.
- Build a neural network to predict the next player move and evaluate the current position.
- Combine the results of MCTS with the neural network training outcomes.
- Set up the game players to train by playing against each other.
- Implement the Connect Four game interface and teach your game players how to play.
- Implement the Dots and Boxes game interface and teach your game players how to play. Experiment with the web interface to play against the models you have trained.
Prerequisite knowledge:
- Basic understanding of machine learning and deep learning concepts
- Familiarity with Python, NumPy and PyTorch
- Some experience with neural networks is recommended but not required