This document describes how to exactly reproduce the experiments in [Scheier2022]
[Scheier2022] Scheiermann, Johannes; Konen, Wolfgang: „AlphaZero-Inspired Game Learning: Faster Training by Using MCTS Only at Test Time”, IEEE Transactions on Games, 2022, doi:10.1109/TG.2022.3206733, preprint available at https://arxiv.org/abs/2204.13307.
Download GBG from https://github.com/WolfgangKonen/GBG, start or compile it according to https://github.com/WolfgangKonen/GBG/wiki/Install-and-Configure.
If not present, build jartools/GBGBatch.jar.
Syntax:
GBGBatch gameName n agentFile [ nruns maxGameNum csvFile scaPar0 scaPar1 scaPar2 propsName ]
propsName codes the properties file name. If not given, it defaults to props_batch.txt.
For more information on GBGBatch, see the extensive Javadoc in GBGBatch.java.
To run the visualization scripts in R_plotTools/, you need to have R >= 4.0 installed.
The following files are included for reference in this repository. They should be present within the indicated directories of the actual GBG distribution from GitHub and they should be started from there:
- experiments/Othello/experim*.sh
- src/starters/props_batch.txt
- src/starters/GBGBatch.java
- resources/R_plotTools/*.R
All scripts *.sh should be run from the GBG distribution.
For more information on the shell scripts found in experiments/Othello/, see Papers-with-Code-Othello.pdf.