Learning of Evaluation Functions to Realize Playing Styles in Shogi

Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 9810)

Abstract

This paper presents a method to give a computer player an intended playing style by the machine learning of an evaluation function. Recent improvements in machine learning techniques have realized the automated tuning of the feature weight vector of an evaluation function. To make a strong player, as many moves as possible of strong players’ game records are needed, though the number of available game records decreases when we focus on a specific playing style. To pursue both goals of playing style and playing strength, we present three steps of learning: classifying moves with respect to playing styles, training the weight vector of an evaluation function by using the whole set of game records to maximize its playing strength, and modifying the weight vector carefully so as to improve agreement with the moves of the intended playing style. We applied our method to realize players of defense or attack-oriented style in shogi and tested the players by self-play against the original version. The results confirmed that the presented method successfully adjusted evaluation functions in that the frequency of defensive moves is significantly increased or decreased in accordance with the game records used while keeping the winning ratio at almost 50 %.

References

  1. Baxter, J., Tridgell, A., Weaver, L.: Learning to play chess using temporal-differences. Mach. Learn. 40(3), 242–263 (2000)CrossRefMATHGoogle Scholar
  2. Buro, M.: Improving heuristic mini-max search by supervised learning. Artif. Intell. 134(1–2), 85–99 (2002)CrossRefMATHGoogle Scholar
  3. Campbell, M., Hoane Jr., A.J., Hsu, F.-H.: Deep blue. Artif. Intell. 134(1–2), 57–83 (2002)CrossRefMATHGoogle Scholar
  4. Carmel, D., Markovitch, S.: Learning models of opponent’s strategy in game playing. In: In Proceedings of the AAAI Fall Symposium on Games: Planning and Learning, pp. 140–147. The AAAI Press (1993)Google Scholar
  5. Donkers, H., van den Herik, H., Uiterwijk, J.: Selecting evaluation functions in opponent-model search. Theoret. Comput. Sci. 349(2), 245–267 (2005)MathSciNetCrossRefMATHGoogle Scholar
  6. Hoki, K., Kaneko, T.: Large-scale optimization for evaluation functions with minimax search. J. Artif. Intell. Res. (JAIR) 49, 527–568 (2014)MathSciNetMATHGoogle Scholar
  7. Hoki, K., Muramatsu, M.: Efficiency of three forward-pruning techniques in shogi: futility pruning, null-move pruning, and late move reduction (LMR). Entertain. Comput. 3(3), 51–57 (2012)CrossRefGoogle Scholar
  8. Iida, H., Sakuta, M., Rollason, J.: Computer shogi. Artif. Intell. 134(1–2), 121–144 (2002)CrossRefMATHGoogle Scholar
  9. Kaneko, T.: Evaluation functions of computer shogi programs and supervised learning using game records. J. Jpn. Soc. Artif. Intell. 27(1), 75–82 (2012). (In Japanese)Google Scholar
  10. Levene, M., Fenner, T.I.: A methodology for learning players’ styles from game records. In: CoRR abs/0904.2595 (2009)Google Scholar
  11. Matsubara, H.: Declaration of termination of computer shogi project. IPSJ Mag. 56(11), 1054–1055 (2015). (In Japanese)Google Scholar
  12. Namai, S., Ito, T.: A trial AI system with its suggestion of Kifuu (playing style) in shogi. In: 2010 International Conference on Technologies and Applications of Artificial Intelligence (TAAI), pp. 433–439 (2010). doi:10.1109/TAAI.2010.94
  13. Okamoto, K., Hashiguchi, H.: Psychological Analysis by Rorschach, MDS of 11 professional shogi players’ Kifuu. Brain Shuppan. (In Japanese)Google Scholar
  14. Ortiz B., S.E., et al.: An interesting opponent for fighting videogames. In: SIG Technical Reports. GI 4. IPSJ, pp. 1–8 (2010)Google Scholar
  15. Sawa, N., Ito, T.: Statistical analysis of elements of play style in shogi (Japanese Chess). In: SIG Technical Reports. GI 3. IPSJ, pp. 1–8 (2011). (In Japanese)Google Scholar
  16. Shimizu, S., Kaneko, T.: Evaluation, implementation of UCT with prior knowledge for computer’s styles of playing two-player games. In: Proceedings of 19th Game Programming Workshop 2014, pp. 188–195 (2014). (In Japanese)Google Scholar
  17. Takise, R., Tanaka, T.: Development of entering-king oriented shogi programs. IPSJ J. 53(11), 2544–2551 (2012). (In Japanese)Google Scholar
  18. Tesauro, G.: Programming backgammon using self-teaching neural nets. Artif. Intell. 134(1–2), 181–199 (2002)CrossRefMATHGoogle Scholar
  19. Tosaka, K., Matsubara, H.: Feature extraction of players from game records in shogi. In: SIG Technical Reports. 2006-GI-016. IPSJ, pp. 1–8 (2006). (In Japanese)Google Scholar
  20. Veness, J., et al.: Bootstrapping from game tree search. Adv. Neural Inf. Process. Syst. 22, 1937–1945 (2009)Google Scholar
  21. Yano, Y., et al.: Adaptive learning utilizing parameters of existing evaluation function. In: Proceedings of 14th Game Programming Workshop 2009, pp. 1–8 (2009). (In Japanese)Google Scholar

Copyright information

© Springer International Publishing Switzerland 2016

Authors and Affiliations

  1. 1.Graduate School of Arts and SciencesThe University of TokyoTokyoJapan

Personalised recommendations