An Improved Population-Based Incremental Learning Method for Inverse Problems

摘要

To eliminate the use of complex operators such as crossover and mutation ones in genetic algorithms, attempts are made to exploit the use of population based incremental learning (PBIL) evolution algorithm. In order to preserve the global search ability without destroying the conceptual simplicity and implementation easiness of existing PBIL methods, some improvements are proposed in this paper. The major enhancements include the use of the best solution in the current iterative cycle to guide the search towards promising regions, the employment of the worst solution to shift the search away from the worst individuals to speed up the convergence, and the introduction of a novel mechanism for generating new individuals to improve the global search ability. To validate and illustrate the merits of the proposed PBIL algorithm, two different optimal design problems are studied, and the numerical results are reported and compared in the paper.