Optimistic Selection Rule Better Than Majority Voting System
Abstract
A recently proposed ensemble approach to game-tree search has attracted a great deal of attention. The ensemble system consists of M computer players, where each player uses a different series of pseudo-random numbers. A combination of multiple players under the majority voting system would improve the performance of a Shogi-playing computer. We present a new strategy of move selection based on the search values of a number of players. The move decision is made by selecting one player from all M players. Each move is selected by referring to the evaluation value of the tree search of each player. The performance and mechanism of the strategy are examined. We show that the optimistic selection rule, which selects the player that yields the highest evaluation value, outperforms the majority voting system. By grouping 16 or more computer players straightforwardly, the winning rates of the strongest Shogi programs increase from 50 to 60% or even higher.
Preview
Unable to display preview. Download preview PDF.
References
- 1.Knuth, D.E., Moore, R.W.: An Analysis of Alpha-Beta Pruning. Artificial Inteligence 6, 293–326 (1975)MathSciNetCrossRefMATHGoogle Scholar
- 2.Slate, D.J., Atkin, L.R.: Chess 4.5 – the northwestern university chess program. In: Frey, P.W. (ed.) Chess Skill in Man and Machine, pp. 82–118. Springer, Heidelberg (1977)Google Scholar
- 3.Moore, G.E.: Cramming More Components onto Integrated Circuits. Electronics 38, 114–117 (1965)Google Scholar
- 4.Hyatt, R.: The DTS High-performance Parallel Tree Search Algorithm, http://www.cis.uab.edu/info/faculty/hyatt/search.html
- 5.Hsu, F.-H.: IBS’s Deep Blue Chess Grandmaster Chips. IEEE Micro 19, 70–80 (1999)CrossRefGoogle Scholar
- 6.Donninger, C., Lorenz, U.: The Chess Monster Hydra. In: Becker, J., Platzner, M., Vernalde, S. (eds.) FPL 2004. LNCS, vol. 3203, pp. 927–932. Springer, Heidelberg (2004)CrossRefGoogle Scholar
- 7.Newborn, M.M.: Unsynchronized Iteratively Deepening Parallel Alpha-Beta Search. IEEE Transactions on Pattern Analysis and Machine Intelligence PAMI-10, 687–694 (1988)CrossRefMATHGoogle Scholar
- 8.Vučković, V.V.: The Method of the Chess Search Algorithms Parallelization using Two-processor Distributed System. In: Facta Universitatis. Series Mathematics an Informatics, vol. 22, pp. 175–188 (2007)Google Scholar
- 9.Brockington, M.G., Schaeffer, J.: The APHID Parallel alpha-beta Search Algorithm. In: IEEE Symposium of Parallel and Distributed Processing (SPDP 1996), New Orleans, October 23–26 (1996)Google Scholar
- 10.Himstedt, K., Lorenz, U., Möller, D.P.F.: A twofold distributed game-tree search approach using interconnected clusters. In: Luque, E., Margalef, T., Benítez, D. (eds.) Euro-Par 2008. LNCS, vol. 5168, pp. 587–598. Springer, Heidelberg (2008)CrossRefGoogle Scholar
- 11.Polikar, R.: Ensemble Based Systems in Decision Making. IEEE Circuits and systems magazine 6, 21–45 (2006)CrossRefGoogle Scholar
- 12.Hanawa, M., Ito, T.: The Council System in Brain Game. In: Cognitive Science and Entertainment Symposium, The University of Electro-Communications, Tokyo, Japan, pp. 72–75 (2009) (in Japanese) Google Scholar
- 13.Obata, T., Hanawa, M., Ito, T.: Consultation Algorithm in Brain Game: Performance of the Simple Majority System. In: Proceedings of the GI22 Conference of IPSJ, pp. 1–5 (2009) (in Japanese) Google Scholar
- 14.Ito, T.: Consultation Player “Monju” – A Majority Rule to Increasing the Winning Rate. IPSJ Journal 50, 887–894 (2009) (in Japanese) Google Scholar
- 15.Grimbergen, R.: The 19th CSA World Computer-Shogi Championship. ICGA Journal 32, 121–125 (2009)CrossRefGoogle Scholar
- 16.Bonanza 4.1.2 is a computer Shogi program. Source codes are available from, http://www.geocities.jp/bonanza_shogi/ (in Japanese)
- 17.GPS Shogi is the first-prize winner of the 19th CSA World Computer-Shogi Championship, Source codes are available from, http://gps.tanaka.ecc.u-tokyo.ac.jp/gpsshogi/ (in Japanese)
- 18.Iida, H., Sakuta, M., Rollason, J.: Computer Shogi. Artificial Intelligence 134, 121–144 (2002)CrossRefMATHGoogle Scholar