This repository contains the code used to produce the results of the experiments in my master thesis.
- Rust 1.38
- Python 3
- tensorflow 1.14
- stable-baselines 2.6
- gym 0.15
Below is an unordered table of experiments and their code.
Experiment | Code |
---|---|
Branching Factor (3.2) | src/plot/possible_plays/mod.rs |
First Player Advantage (3.3) | src/game_player/advantage_checker.rs |
Transposition Table Benchmark (4.1.5) | benches/game_logic/transposition.rs |
Performance Rust (5.1.1) | benches/game_logic/field.rs |
Human Play web interface (6.2.1) | Web Version and the experiment code part |
Reproducibility (6.3) | python-rl/repro.py and scripts/reprodocubilityDataProcessor.py |
Observation Space (6.4) | python-rl/searchparams.py |
Reward Function Effectiveness (6.5) | scripts/reprodocubilityDataProcessor.py |
MCTS Effectiveness (6.6) | src/game_player/mctsparameval.rs |
RL Agents Training (6.7) | python-rl/selfplay.rs |
Tournament (6.8) | src/game_player/tournament.rs |
MCTS Verification (6.9) | src/hex/hex_eval.rs |
Untested and unused AlphaZero Rust pytorch implementation | src/rl/aznet.rs |