Computing at Glasgow University
Paper ID: 8206

Failing First: An Update
Beck,J.C. Prosser,P. Wallace,R.J.

Publication Type: Conference Proceedings
Appeared in: 16th European Conference on Artificial Intelligence
Page Numbers :
Publisher: N/A
Year: 2004

In ECAI 1998 Smith & Grant performed a study of the fail-first principle of Haralick & Elliott. The fail-first principle states that "To succeed, try first where you are most likely to fail." The basic hypothesis of Smith & Grant was that if failing first is such a good thing then more of it must be better. Therefore, creating heuristics with a stronger ability to fail early should increase search efficiency. Their experiments showed that this was not the case. We repeated those experiments but did not replicate their results (and this was due to a bug in their code). We draw a more specific conclusion: trying harder to fail-first does indeed reduce search effort but this phenomenon is algorithm dependent

Keywords: constraint satisfaction fail-first heuristic

PS/PS.GZ PDF Bibtex entry Endnote XML