A data-based coding of candidate strings in the closest string problem

Given a set of strings S of equal lengths over an alphabet Σ, the closest string problem seeks a string over Σ whose maximum Hamming distance to any of the given strings is as small as possible. A data-based coding of strings for evolutionary search represents candidate closest strings as sequences of indexes of the given strings. The string such a chromosome represents consists of the symbols in the corresponding positions of the indexed strings. A genetic algorithm using this coding was compared with two GAs that encoded candidate strings directly as strings over Σ. In trials on twenty-five instances of the closest string problem with alphabets ranging is size from 2 to 30, the algorithm that used the data-based representation of candidate strings consistently returned the best results, and its advantage increased with the sizes of the test instances’ alphabets. In Proceedings of the Genetic and Evolutionary Computation Conference, GECCO–2009 (G¨unther Raidl et al., Editors...

• Type: Conference
• Year: 2009
• Where: GECCO
• Authors: Bryant A. Julstrom
• Keywords: Candidate Closest Strings, Closest String, Closest String Problem, GECCO 2009, Optimization