Gomoku-AI is a personal learning project to
- build a comand-line-based Gomoku-like game (multiplayer m,n,k-game) with human as well as computer players, and around it
- create an algorithm development platform for various game board algorithms including (hopefully) machine learning ones. It also serves to
- gather experience on larger-scale software development, version control and project management.
Roadmap (summary)
-
2021-10: Tic Tac Toe
- 3x3 board, 2 players (optionally computer players)
- Computer players use the minimax algorithm. They cannot lose.
- Internally: Now supports many algorithms and game variations.
-
2021-06: Core Game
- Text-based Gomoku-like game. Played by inputting coordinates.
- 30x30 board, 3 human players
- Internally: Functionality neatly separated into classes, tests for all features
- Limited-depth minimax with scoring function
- Core game expansion with computer players and new algorithm
- Gradient descent search for optimal scoring function
- Alpha-beta-pruning algorithm
- Remembering already calculated moves
- Real machine learning algorithms