Releases: masus04/Deep-Reinforcement-Learning-for-Boardgames
Releases · masus04/Deep-Reinforcement-Learning-for-Boardgames
Unified Experiments
0.65 Preparation
Major network updates
- Make use of LogSoftmax().exp() for numerically stable and non spiking LegalSoftmax module
- Fixes TTT Baseline Player's loss function bug
- Improved plotting
Major TicTacToe fix
Resolved config file issue which impacts all TicTacToe experiments. All TicTacToe experiments are run after this point.
Major Functionality release
Includes major functionality:
- Framework
- TicTacToe & Othello
- Reinforce, Baseline & Actor Critic players
- Search player
- GUI
-> All experiments are based on this release. Will create new release if learning players are changed.