Olivier Teytaud

Genetic fuzzy markup language for game of NoGo

NoGo is similar to the game of Go in terms of gameplay; however, the goal is different: the first player who either suicides or kills a group loses the game and the first player with no legal move loses the game. In this paper, we propose an approach combining the technologies of ontologies, evolutionary computation, fuzzy logic, and fuzzy markup language (FML) with a genetic algorithm (GA)-based system for the NoGo game. Based on the collected patterns and the pre-constructed fuzzy NoGo ontology, the genetic FML (GFML) with the fuzzy inference mechanism is able to analyze the situation of the current game board and then play next move to an inferred good-move position. Additionally, the genetic learning mechanism continuously evolves the adopted GFMLs to enable an increase in the winning rate of the GA-based NoGo via playing with the baseline NoGo. In the proposed approach, first, the domain experts construct the important NoGo patterns and the fuzzy NoGo ontology based on the rules of NoGo and the past game records. Second, each GA-based NoGo as White plays against the baseline NoGo as Black according to the inferred and calculated good-move position, respectively. Third, the genetic learning mechanism is carried out to generate two new evolved GFMLs and then the worst two GFMLs stored in the GFML repository are replaced. Fourth, the GFML with the highest winning rate is randomly sampled from the GFML repository in the time series. Finally, one by one the GA-based NoGo adopts the sampled GFML to play lots of games against the baseline NoGo to obtain the winning rate of the GA-based NoGo. The acquired winning rates at the time series show that the proposed approach can work effectively and that the average winning rate of the GA-based NoGo program is much stronger than the baseline NoGo program.

Lemmas on partial observation, with application to phantom games

Solving games is usual in the fully observable case. The partially observable case is much more difficult; whenever the number of strategies is finite (which is not necessarily the case, even when the state space is finite), the main tool for the exact solving is the construction of the full matrix game and its solving by linear programming. We here propose tools for approximating the value of partially observable games. The lemmas are relatively general, and we apply them for deriving rigorous bounds on the Nash equilibrium of phantom-tic-tac-toe and phantom-Go.

Fuzzy Ontologies for the Game of Go

This chapter presents a developed fuzzy ontology model for computer Go applications. Unlike previous research, this chapter employs features derived from professional Go players’ domain knowledge to transform them into the opening book sequence and to represent them by a fuzzy ontology for the game of Go. Afterward, the domain experts validate the built fuzzy ontology. The developed fuzzy ontology has been verified through the invited games for Go programs playing against human Go players. The results show that the fuzzy ontology can work for computer Go application.

Upper confidence tree-based consistent reactive planning application to minesweeper

Many reactive planning tasks are tackled through myopic optimization-based approaches. Specifically, the problem is simplified by only considering the observations available at the current time step and an estimate of the future system behavior; the optimal decision on the basis of this information is computed and the simplified problem description is updated on the basis of the new observations available in each time step. While this approach does not yield optimal strategies stricto sensu, it indeed gives good results at a reasonable computational cost for highly intractable problems, whenever fast off-the-shelf solvers are available for the simplified problem. n nThe increase of available computational power − even though the search for optimal strategies remains intractable with brute-force approaches − makes it however possible to go beyond the intrinsic limitations of myopic reactive planning approaches. n nA consistent reactive planning approach is proposed in this paper, embedding a solver with an Upper Confidence Tree algorithm. While the solver is used to yield a consistent estimate of the belief state, the UCT exploits this estimate (both in the tree nodes and through the Monte-Carlo simulator) to achieve an asymptotically optimal policy. The paper shows the consistency of the proposed Upper Confidence Tree-based Consistent Reactive Planning algorithm and presents a proof of principle of its performance on a classical success of the myopic approach, the MineSweeper game.