Linkage tree genetic algorithm: first results

摘要: We introduce the Linkage Tree Genetic Algorithm (LTGA), a competent genetic algorithm that learns linkage between problem variables. The LTGA builds each generation tree using hierarchical clustering algorithm. To generate new offspring solutions, selects two parent solutions and traverses starting from root. At branching point, pair is recombined crossover mask represented by clusters are merged at particular node. competes with pair, continues traversing has most fit solution. Once entire traversed, best solution of current copied to next generation. In this paper we use normalized variation information metric as distance measure for process. Experimental results classical fully deceptive function show only requires very small, minimal population sizes, executes similar number evaluations existing learning algorithms.