Improved Adaptive Differential Evolution Algorithm And Its Applications

In recent years, as a new subject of computational intelligence, evolutionary com-putation has developed rapidly. Evolutionary algorithm contains some branches, such as Genetic Algorithm, Genetic Programming, Evolution strategics, Evolutionary Pro-gramming, Differential Evolution; and these algorithms have not only some different features but also some common properties. Since Evolutionary algorithms are self-adaptive, auto-optimizing, parallel, it is reasonable and available to solve the problems in the science and engineering by using evolutionary algorithms.Differential Evolution (DE) is a kind of evolutionary algorithms. It has received wide attention for its simple structure, fast convergence, and robustness. In particular, the differential evolution algorithm is an efficient population-based search algorithm for global optimization. There are three main control parameters of the DE algorithm: the mutation scale factor, the crossover constant, and the population size. These pa-rameters are of great importance to the efficiency of a DE algorithm. DE algorithm has been successfully applied in diverse fields such as data mining, pattern recognition, digital filters, artificial neural network, combinatorial optimization, multi-objective op-timization, etc. However, the original DE algorithm has some shortages, so it is urgent to improve the original DE algorithm.In this thesis, the differential evolution algorithm and its application are investi-gated. To begin with, the importance of the control parameters for differential evolution algorithm is given, and the population adaptive parameter strategy is proposed. Then for discrete space optimization, we introduce the adaptive parameter of binary differ-ential evolution algorithm. Furthermore, the multi-objective self-adaptive differential evolution algorithm is designed to solve numerical optimization problems with mul-tiple conflicting objectives. In addition, the SAPA algorithm is used to solve global optimization problems with applications in identifying unknown parameters of a class of genetic regulatory networks (GRNs) with random delays and stochastic perturba-tions. Finally, the intelligent partition for3-D basic body surface is studied by the DE algorithm. The main contents and the innovative points can be listed as follows: (1) A self-adaptive DE with population adjustment scheme. A self-adaptive DE with population adjustment scheme (SAPA) is proposed to tune the size of offspring population. The novel algorithm involves two self-adaptive DE strategies and two population adjustment schemes. The performance of the SAPA algorithm is evaluated by a set of benchmark functions. Simulation results show that the proposed algorithm is better than, or at least comparable to, other classic or adaptive DE algorithms. Performance comparisons with some other well-known evolutionary algorithms from literatures are also presented.(2)Population Adaptive Binary Differential Evolution Algorithm. By improving the mutation of the DE/current-to-best/1strategy, we propose a new pop-ulation adaptive binary differential Evolution, which can be applied to the discrete optimization space. The proposed algorithm adaptive adjusts in accordance with the solution-searching status, which improves the efficiency of the algorithm and optimized accuracy. The experimental comparison with other binary DE algorithm indicates that the population adaptive binary differential evolution algorithm have better convergence and higher accuracy.(3)Multi-objective Self-adaptive Differential Evolution Algorithm. Multi-objective Self-adaptive Differential Evolution (MOSDE) algorithm is proposed to solve numerical optimization problems with multiple conflicting objectives. The Multi-objective Self-adaptive Differential Evolution algorithm involves three self-adaptive DE strategies. The usefulness of the MOSDE algorithm is demonstrated with extensive nu-merical experiments showing improvements in performance compared with the previous state of the art.(4)Parameter identification of stochastic genetic regulatory networks with random delays. Based on the differential evolution algorithm, the parameter identification of stochastic genetic regulatory networks with random delays is studied. The simulation results show that SAPA algorithm is superior or comparable to the other algorithms and can be efficiently used for identifying the unknown parameters of stochastic genetic, regulatory networks with random delays.(5)The Intelligent Partition for3-D Basic Body Surface Based on Dif-ferential Evolution. Through analyzing the characteristics of the3-D Basic Body Surface, the3-D surface flattening and segmentation problem are studied. Then the intelligent partition algorithm based on differential evolution is designed, which solves the problem of partition of basic body surface. Simulation results show that the pro-posed algorithm is better than, or at least comparable to, other Evolutionary algorithm and can effectively search the optimum route which segments the basic body surface.