Dual lookups in pattern databases

Ariel Felner, Uzi Zahavi, Jonathan Schaeffer, Robert C. Holte

Research output: Contribution to journalConference articlepeer-review

52 Scopus citations

Abstract

A pattern database (PDB) is a heuristic function stored as a lookup table. Symmetries of a state space are often used to enable multiple values to be looked up in a PDB for a given state. This paper introduces an additional PDB lookup, called the dual PDB lookup. A dual PDB lookup is always admissible but can return inconsistent values. The paper also presents an extension of the well-known pathmax method so that inconsistencies in heuristic values are propagated in both directions (child-to-parent, and parent-to-child) in the search tree. Experiments show that the addition of dual lookups and bidirectional pathmax propagation can reduce the number of nodes generated by IDA* by over one order of magnitude in the TopSpin puzzle and Rubik's Cube, and by about a factor of two for the sliding tile puzzles.

Original languageEnglish
Pages (from-to)103-108
Number of pages6
JournalIJCAI International Joint Conference on Artificial Intelligence
StatePublished - 1 Dec 2005
Event19th International Joint Conference on Artificial Intelligence, IJCAI 2005 - Edinburgh, United Kingdom
Duration: 30 Jul 20055 Aug 2005

Fingerprint

Dive into the research topics of 'Dual lookups in pattern databases'. Together they form a unique fingerprint.

Cite this