×

The comparison between the statistical heuristic search and \(A^*\). (English) Zbl 0719.62093

Assuming G is a uniform m-ary tree, where m is its branching factor, the heuristic statistical search algorithm (SA) consists of the statistical inference method SPRT (Wald sequential probability ratio test) and BF (best-first) heuristic search. Then the procedure of SA search is divided into two steps: First, it quickly identifies the most promising subtree using a subtree evaluation function a(n) (global statistic or subtree statistic). Based on SPRT, from nodes \(n_{11},n_{12},...,n_{1m}\) in the first level, the m search directions, i.e., the subtrees \(T(n_{11}),...,T(n_{1m})\) rooted at nodes \(n_{11},...,n_{1m}\), are selected and rejected, taking the expanded nodes in each subtree as observed samples. The subtrees which contain the goal with lower probability are rejected. The most promising one is selected. Then, it expands nodes within that subtree using a node evaluation function b(n) (local statistic or node statistic).

MSC:

62L99 Sequential statistical methods
68R99 Discrete mathematics in relation to computer science
62L10 Sequential statistical analysis
05C05 Trees
PDFBibTeX XMLCite
Full Text: DOI

References:

[1] Zhang Bo and Zhang Ling. Statistical heuristic search.J. of Compat. Sci. & Technol.,2:1 (1987), 1–11. · Zbl 0641.68088
[2] G. B. Wetherill, Sequential Method in Statistics, John Wiley & Sons, Inc., (1975). · Zbl 0343.62073
[3] Shelemyahu Zacks, The Theory of Statistical Inference, John Wiley & Sons, Inc., (1971). · Zbl 0321.62003
[4] J. Pearl, Heuristic search theory: survey of recent results, in Proc. 7th IJCAI-81, 554–562.
[5] J. Pearl, Heuristics, Addison–Wesley Publishing Company (1984).
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.