Chapter 8 - Stochastic search methods for CSPs

Abstract

Stochastic search is a class of search methods that includes heuristics and an element of nondeterminism in traversing the search space. This chapter introduces two stochastic search methods, one is based on hill-climbing and the other based on a connectionist approach. Both of them are general techniques that have been used in problems other than the constraint satisfaction problems (CSPs). The chapter focuses on their application in CSP solving. Hill-climbing is a general search technique that has been used in many areas, for example, optimization problems, such as the well known Travelling Salesman Problem. Recently, it has been found that hill-climbing using the min-conflict heuristic can be used to solve the N-queens problem more quickly than other search algorithms. The connectionist approaches to CSPs have attracted great attention because of their potential for massive parallelism that gives hope to the solving of problems that are intractable under conventional methods or of solving problems with a fraction of the time required by conventional methods, sometimes at the price of losing completeness. GENET is a connectionist model for CSP solving. It has demonstrated its effectiveness in binary constraint problems and is being extended to tackle general CSPs.