Search space preprocessing in solving complex optimization problems

We look at the complexity placed by big search spaces, dominated by the number of variables and domain of each variable, in search and optimization problems. While a large, even infinite, search domain impairs the effectiveness and efficiency of search, a complex structure of constraints further increases the difficulty in that the search space becomes highly irregular. We propose in this position paper that data mining and dimension reduction techniques have a potential in addressing the pressing issues in both combinatorial optimization and continuous optimization. By preprocessing the original search space, data mining can help boost the speed of search by guiding the search effort to a reduced, more promising area.