An Experimental Analysis of the B* Tree Search Algorithm
Abstract
A set of selection rules is presented for guiding the B* search procedure. The selection rules are based on a simple probability model. Using those selection procedures, the B* algorithm will expand approximately one-third as many nodes as a best first search and approximately two-thirds as many nodes as the previous best version of the B* algorithm. The B* algorithm, using the selection procedures, will expand 30% more nodes than optimum, for small trees, and up to 170% for larger trees. The issues of when the B* paradigm is useful is examined with the effect of invalid bounds on the algorithm.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 08, 1980
- Accession Number
- ADA224252
Entities
People
- Andrew J. Palay
Organizations
- Carnegie Mellon University