We introduce a distinction, in single-agent problems, between transpositions that are due to permutations of commutative moves and transpositions that are not. We show a simple modification of a depth-first search algorithm which can detect the transpositions of the first class without the use of a transposition table. It works by maintaining, for each node, a list of moves that are known to lead to transpositions. This algorithm is applied to two one-player games: a solitary card game called Gaps, and a game called Morpion Solitaire. We analyze, for each domain, how often transpositions are due to commutative moves. In one variant of Gaps, the algorithm enables to search more efficiently with a small transposition table. In Morpion Solitaire, a transposition table is not even needed. The best known sequence for this game is proved optimal for more than one hundred moves.
[1]
Wolfgang Banzhaf,et al.
Advances in Artificial Life
,
2003,
Lecture Notes in Computer Science.
[2]
Tristan Cazenave,et al.
Searching with Analysis of Dependencies in a Solitaire Card Game
,
2003,
ACG.
[3]
Hugues Juillé.
Incremental Co-Evolution of Organisms: A New Approach for Optimization and Discovery of Strategies
,
1995,
ECAL.
[4]
Jordan B. Pollack,et al.
Methods for statistical inference: extending the evolutionary computation paradigm
,
1999
.
[5]
D. M. Breuker.
Memory versus search in games
,
1998
.