Constraint-Based Symmetry Detection in General Game Playing
Authors: Frédéric Koriche, Sylvain Lagrue, Éric Piette, Sébastien Tabary
IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Based on a theoretical analysis of this approach, we experimentally show on various games that the recent stochastic constraint solver MAC-UCB, coupled with constraint-based symmetry detection, significantly outperforms the standard Monte Carlo Tree Search algorithms, coupled with rule-based symmetry detection. This constraint-driven approach is also validated by the excellent results obtained by our player during the last GGP competition. |
| Researcher Affiliation | Academia | Fr ed eric Koriche, Sylvain Lagrue, Eric Piette and S ebastien Tabary CRIL, CNRS UMR 8188, Universit e d Artois, France {koriche, lagrue, epiette, tabary}@cril.fr |
| Pseudocode | No | The paper describes algorithms like MAC-UCB, UCT, and GRAVE, but it does not include any structured pseudocode or algorithm blocks for its own proposed method. |
| Open Source Code | No | The paper does not provide any explicit statement or link indicating that the source code for their described methodology is open-source or publicly available. |
| Open Datasets | No | The paper mentions testing on '20 deterministic GDL games, and 5 stochastic GDL games (with random)', including specific games like Tic-Tac-Toe and Chess. However, these are general game types, not formal datasets with citations, DOIs, or repository links for public access. |
| Dataset Splits | No | The paper specifies the duration for 'deliberation time before the first turn' (180s) and 'deliberation time per turn' (15s) and the number of game contests. However, it does not describe specific training, validation, or test dataset splits in the conventional sense for model development. |
| Hardware Specification | Yes | Based on our framework, we now present a series of experiments conducted on a cluster of Intel Xeon E5-2643 CPU 3.3 GHz with 64 GB of RAM and four threads under Linux. |
| Software Dependencies | No | The paper mentions using the 'NAUTY algorithm [Mc Kay and Piperno, 2014]' and a 'hash function specified in [Zobrist, 1990]', but it does not provide specific version numbers for these or any other software components. |
| Experiment Setup | Yes | The horizon T was fixed to 200. The choice of the depth d was derived by partitioning the deliberation time into three ratios (r MAC, r UCB, r SYM), which correspond to the proportions of runtime allocated for MAC, FLAT-UCB, and symmetry detection, respectively. Based on a sensitivity analysis of MAC-UCB-SYM, we used the ratios (45%, 30%, 25%). MAC-UCB-SYM uses a hash table of 32Gb with a hash function specified in [Zobrist, 1990]. |