New adaptive search methods based on multiresolution analysis and wavelet theory are introduced and discussed within the framework of Markov theory. These stochastic search methods are suited to problems for which good solutions tend to cluster within the search space. Multiresolution search methods are extended to searches with memory. The introduction of a memory allows an easy inclusion of local information available prior to the search and the storage of a low resolution approximation of the fitness function. Further, by using B-splines, a linguistic, fuzzy interpretation of the search results can be given. The relation between wavelet-based search methods and wavelet estimation theory is explained.
展开▼
