Abstract
The ancient oriental game of Go is the last of the great perfect knowledge games wherein strong human players still outperform AI players, but the gap is narrowing. Superior strategic skill is keeping humans several steps ahead even as computers catch up tactically. Taking the perspective that large scale topological perception is part of the key to strategic excellence, we present a new method for generating strategic plans in the game of Go. Treating stones in play as continuous perturbations of a dynamical system composed of coupled non-linear oscillators, we exploit transitory synchrony in the wave patterns thus generated to make topological inferences about possible territories later in the game. Our system is shown, using professional game records, to focus search on appropriate areas of the space of continuations. By combining spatial (rather than game-tree) representation with dynamical lookahead, we hope to overcome some of the limitations faced by static spatial mapping and dynamical game-tree based approaches.
Original language | English |
---|---|
Title of host publication | ECAI 2010 |
Subtitle of host publication | 19th proceedings |
Place of Publication | Amsterdam |
Publisher | IOS Press |
Pages | 1095-1096 |
Number of pages | 2 |
ISBN (Electronic) | 9781607506058 |
Publication status | Published - 2010 |
Event | European Conference on Artificial Intelligence - Lisbon, Portugal, Portugal Duration: 16 Aug 2010 → 20 Aug 2010 |
Conference
Conference | European Conference on Artificial Intelligence |
---|---|
Country/Territory | Portugal |
Period | 16/08/10 → 20/08/10 |