Grid coevolution for adaptive simulations; application to the building of opening books in the game of Go

Pierre Audouard (1) , Guillaume Chaslot (2) , Jean-Baptiste Hoock (3) , Arpad Rimmel (3) , J. Perez (3) , Olivier Teytaud (3, 4, 5)

1 Go expert
2 Maastricht University [Maastricht]
3 TAO - Machine Learning and Optimisation
4 TANC - Algorithmic number theory for cryptology
5 LRI - Laboratoire de Recherche en Informatique

This paper presents a successful application of parallel (grid) coevolution applied to the building of an opening book (OB) in 9x9 Go. Known sayings around the game of Go are refound by the algorithm, and the resulting program was also able to credibly comment openings in professional games of 9x9 Go. Interestingly, beyond the application to the game of Go, our algorithm can be seen as a ”meta”-level for the UCT-algorithm: ”UCT applied to UCT” (instead of ”UCT applied to a random player” as usual), in order to build an OB. It is generic and could be applied as well for analyzing a given situation of a Markov Decision Process.

Domaines: Optimisation et contrôle [math.OC]

Pierre Audouard, Guillaume Chaslot, Jean-Baptiste Hoock, Arpad Rimmel, J. Perez, et al.. Grid coevolution for adaptive simulations; application to the building of opening books in the game of Go. EvoGames, 2009, Tuebingen, Germany. ⟨inria-00369783⟩