SuperGo is a student project implementing DeepMind's AlphaGo Zero algorithm for the game of Go. It aims to provide a functional, albeit simplified, version of the system for educational purposes, allowing users to explore self-play, training, and evaluation of Go-playing neural networks.

How It Works

The project utilizes Monte Carlo Tree Search (MCTS) combined with a deep neural network. The network predicts move probabilities and game outcomes, guiding the MCTS. Key features include Dirichlet noise for exploration, adaptive temperature for move selection, and data augmentation via the dihedral group of board symmetries. Training is performed on self-play games, with an ongoing effort to optimize the process and implement features like learning rate annealing.

Quick Start & Requirements

• Install: Not explicitly detailed, but likely involves pip install -r requirements.txt (assuming a requirements.txt exists).
• Prerequisites: PyTorch, pachi_py, Python. GPU acceleration is implied for training and evaluation.
• Resources: Requires significant computational resources for training and self-play, especially for larger board sizes (19x19).

Highlighted Details

• Implements core AlphaGo Zero concepts like MCTS, self-play, and neural network training.
• Utilizes PyTorch for neural network implementation.
• Supports data augmentation through board rotations and reflections.
• Includes features like adaptive temperature and Dirichlet noise for improved search.

Maintenance & Community

This is described as an "ongoing project" by a student. There are no explicit mentions of community channels, significant contributors, or a formal roadmap beyond the TODO list.

Licensing & Compatibility

The license is not specified in the README. Compatibility with commercial or closed-source projects is unknown.

Limitations & Caveats

The project is a student implementation and may not match the performance or robustness of the official AlphaGo Zero. Features like resignation are explicitly noted as missing. The README indicates ongoing development and a significant TODO list, suggesting potential instability or incomplete functionality.