Learning a Go Heuristic with Tilde
Abstract
In Go, an important factor that hinders search is the large branching factor, even in local problems. Human players are strong at recognizing frequently occurring shapes and vital points. This allows them to select the most promising moves and to prune the search tree. In this paper we argue that many of these shapes can be represented as relational concepts. We present an application of the relational learner TILDE in which we learn a heuristic that gives values to candidate-moves in tsume-go (life and death) problems. Such a heuristic can be used to limit the number of evaluated moves. Even if all moves are evaluated, alpha-beta search can be sped up considerably when the candidate-moves are approximately ordered from good to bad.We validate our approach with experiments and analysis.
Keywords
Machine Learning Go Decision trees Inductive Logic Programming Tsume-GoPreview
Unable to display preview. Download preview PDF.
References
- 1.H. Blockeel and L. De Raedt. Top-downinduction of first order logical decision trees. Artificial Intelligence, 101(1-2):285–297, June 1998.MATHCrossRefMathSciNetGoogle Scholar
- 2.H. Blockeel, L. De Raedt, and J. Ramon. Top-down induction of clustering trees. In Proceedings of the 15th International Conference on Machine Learning, pages 55–63, 1998. http://www.cs.kuleuven.ac.be/~ml/PS/ML98-56.ps.
- 3.H. Blockeel, B. Demoen, L. Dehaspe, G. Janssens, J. Ramon, and H. Vandecasteele. Executing query packs in ILP. In J. Cussens and A. Frisch, editors, Proceedings of the 10th International Conference in Inductive Logic Programming, Lecture Notes in Artificial Intelligence, London, UK, July 2000. Springer.Google Scholar
- 4.L. De Raedt. Logical settings for concept learning. Artificial Intelligence,95:187–201, 1997.MATHCrossRefMathSciNetGoogle Scholar
- 5.L. De Raedt. Attribute-value learning versus inductive logic programming: the missing links (extended abstract). In D. Page, editor, Proceedings of the Eighth International Conference on Inductive Logic Programming, volume 1446 of Lecture Notes in Artificial Intelligence, pages 1–8. Springer-Verlag, 1998.Google Scholar
- 6.D. Dyer. An eye shape library for computer go. http://www.andromeda.com/people/ddyer/go/shape-library.html.
- 7.M. Enzenberger. The integration of a priori knowledge into a go playing neural network, 1996. http://home.t-online.de/home/markus.enzenberger/neurogo.html.
- 8.T. Kojima and A. Yoshikawa. A two-step model of pattern acquisition: Application to tsumego. In H. van den Herik and H. Iida, editors, Proceedings of the 1st International Conference on Computers and Games, volume 1558 of Lecture Notes in Computer Science, pages 146–166. Springer-Verlag, 1998.Google Scholar
- 9.S.-H. Nienhuys-Cheng and R. Wolf. Foundations of inductive logic programming, volume 1228 of Lecture Notes in Computer Science and Lecture Notes in Artificial Intelligence. Springer-Verlag, NewYork, NY, USA, 1997.Google Scholar
- 10.F. Provost, T. Fawcett, and R. Kohavi. The case against accuracy estimation for comparing induction algorithms. In Proceedings of the 15th International Conference on Machine Learning, pages 445–453. Morgan Kaufmann, 1998.Google Scholar
- 11.N. Sasaki, Y. Sawada, and J. Yoshimura. Aneural network program of tsume-go. In H. van den Herik and H. Iida, editors, Proceedings of the 1st International Conference on Computers and Games, volume 1558 of Lecture Notes in Computer Science, pages 167–182. Springer-Verlag, 1998.Google Scholar
- 12.D. Stoutamire.Machine learning applied to go. Master’s thesis, Case Western Reserve University, 1991.Google Scholar
- 13.T. Wolf. The program GoTools and its computer-generated tsume go database. Report, School of Mathematical Sciences, Mile End Road London E1 4NS, November 1996. http://www.qmw.ac.uk/~ugah006/gotools/.