An information entropy-based multi-population genetic algorithm
Chunlian Li, Xicheng Wang · Dalian Ligong Daxue xuebao · 2004
An evolutionary design model for constraint optimization problems through transformation of optimization models is constructed, and then a multi-population genetic algorithm with narrowing of the search space is presented to solve the problem. By defining the probabilities that the optimal solution occurs in each population, information-entropy is introduced into evolution process. The probabilities can be obtained by explicitly solving the multi-objective optimization problem with information entropy: they are just Lagrange multipliers in the Kuhn-Tucker condition of the problem, and then the coefficients of narrowing of the searching space for multi-population genetic algorithm can be given by means of them and to control contraction of the solution space. The premature problem can be avoided by keeping diversity among different populations. The algorithm can be ensured by very rapid and steady convergence using solution space contraction criterion. The ability of searching optimization solution for the evolution algorithms is enhanced by introducing entropy. Numerical examples show that the method has very high accuracy and effectiveness.