Split Moves for Monte-Carlo Tree Search

Authors: Jakub Kowalski, Maksymilian Mika, Wojciech Pawlik, Jakub Sutowicz, Marek Szykuła, Mark H. M. Winands10247-10255

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The tests are carried out on a set of board games and performed using the Regular Boardgames General Game Playing formalism, where split strategies of different granularity can be automatically derived based on an abstract description of the game. The results give an overview of the behavior of agents using split design in different ways.
Researcher Affiliation Academia 1University of Wrocław, Faculty of Mathematics and Computer Science 2Maastricht University, Department of Data Science and Knowledge Engineering
Pseudocode Yes Algorithm 1: Vanilla semisplit MCTS.
Open Source Code Yes The full version of this paper is available at (Kowalski et al. 2021b), and the source code used for the experiments is shared within the RBG implementation (Kowalski et al. 2021a).
Open Datasets Yes Our test set consists of 12 board games, well known in GGP (Amazons, Breakthrough, Breakthru, Chess, Chess without check, English Draughts, Fox And Hounds, Go, Knightthrough, Pentago, Skirmish, and The Mill Game).
Dataset Splits No The paper describes testing agents against baselines in timed and fixed settings but does not mention explicit train/validation/test dataset splits like percentages or sample counts for data.
Hardware Specification No The computations were run on a computational grid in the Institute of Computer Science, University of Wrocław, funded by National Science Centre, Poland, under project number 2019/35/B/ST6/04379. (This is a general description of the computing environment, but it lacks specific hardware details like GPU/CPU models, memory, etc.)
Software Dependencies No The paper mentions 'RBG implementation (Kowalski et al. 2021a)' but does not list specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup No All parameters (e.g., the exploration factor, MAST and RAVE policies) were set according to the recommendations in the literature (Finnsson and Björnsson 2010; Sironi and Winands 2016) and keep the same for every agent. (The paper refers to recommendations in literature for parameter settings, but does not state the specific parameter values within the text.)