Evolutionary Strategy Learning, Convergence And Escape Capacity Research And Application

The Lamarckian evolution learning strategy (LELS) and Darwinian evolution learning strategy (DELS) are presented in this paper. Both their similarities and differences in learning implementation are discussed. The former is based on Inheritance of acquired character, i.e., both phenotype and genotype can be optimized through learning. While the later only optimizes phenotype based on Darwinian selection. The convergence of ELS was proved using Markov chain theory. And we also theoretically demonstrated that DELS has stronger escaping capacities. These algorithms are applied to 8 standard test functions. Simulation results show that LELS and DELS yield faster convergence and better global optimization ability than standard evolution strategies: moreover, DELS also leads to better escaping capacity. Finally, the limitations of the work are discussed, as well as the future study.The minimal fuzzy partition, completeness and distinguishability of fuzzy partition, and consistency of rule are defined, with further analysis on Interpretability. In an attempt to develop interpretable fuzzy systems, a systematic design paradigm is proposed using evolution strategy for mixed parameter optimization. The proposed approach is applied to the design of fuzzy controller for a nonlinear plant. It is verified that the interpretable fuzzy system works very well, especially when the priori Knowledge are insufficient.K-means clustering has two disadvantages. one is easily trapped in local minimum. and the other is difficult to determine the number of clusters K. To address the problems, this paper proposes 3 new K-means algorithms based on Evolution Strategy. They own the easier coding scheme and the smaller population than Genetic Algorithm. These algorithms are applied to cluster Fisher's iris data set and work very well, especially when the priori Knowledge are insufficient.