Research On Quantum-inquired And Tabu-based Dierential Evolution Algorithms

Evolutionary algorithms have been successfully applied in many types ofproblems, mainly related to optimization tasks. Di?erential evolution (DE),which is a novel kind of the evolutionary algorithms, has developed fast for 5-7years, and has attracted great attention of a large number of researchers. DEwith the mutation, crossover and selection operators is similar to genetic algo-rithm, genetic programming, evolution strategy and evolution programming.Its mutation operator is the most special operation manner, and its selectionoperator is a"selecting the better"strategy.DE is a population-based optimization technique, which is independentof the problems and can be easily carried out. DE, which was proposed forsolving numerical optimization problems at first, has been extensively appliedto constrained optimization, dynamic optimization, multi-objective optimiza-tion, and so on. Researchers have proposed many modified versions of DEwhich include multi-population, neighboring search, niche, fuzzy adjusting,chaos search, self-adaptive strategy, local search and cooperative search. Inengineering applications, DE has been used in the process control, schedulingproblems, routing problems, knowledge discovery, and filter design.After the current modifications and applications of DE are introduced,the work of the thesis firstly focuses on how to enhance the global search-ing ability of DE. Here, the quantum bits as a representation are used in thequantum-inspired di?erential evolution algorithm. Each individual, which in-cludes more available information, enhances the diversity of the population,and then improve the searching ability. Next, for classification and fuzzy mod-eling, the searching process is accelerated owing to the quantum-inspired dif-ferential evolution algorithm. Thus, the results are improved. The search-ing ability is enhanced for classification and fuzzy model problems by usingthe quantum-inspired di?erential evolution algorithm. Finally, for constrainedproblems with many local minima, the thesis pays attention to how to enhance the ability of escaping from the local minima. Thus the tabu-based di?erentialevolution algorithm is proposed. The main research tasks are summarized asfollows:1. For enhancing the global searching ability, the quantum bit is introducedto propose the quantum-inspired di?erential evolution algorithm (QDE)is proposed. Some search procedures of DE are used to evolve the Q-bit individuals. A new selection operator is introduced to accelerate theconvergence of the Q-bit individuals. Compared to other versions of thebinary di?erential evolution algorithms, the experimental results showthat QDE is excellent. Thus QDE will be a new aspect of the binarydi?erential evolution algorithms.2. In classification problems, the types of the attribute of the data set in-clude the continuous and nominal attributes. For some exiting methods,discretization is necessary to deal with the continuous attribute in thepreprocessing stage. It is proposed a classification algorithm based onDE/QDE. DE/QDE, with the basic procedures of DE, can deal withthe continuous and nominal attributes. Hence, it can easily solve moreclassification problems with mixed attributes.3. For designing T-S fuzzy model, it is proposed two identification method:the one is based on subtractive clustering and DE; the other is based onMDE/QDE. In the identification method based on subtractive cluster-ing and DE, the structure is identified firstly, and then the parametersare calculated. In the identification method based on MDE/QDE, thestructure and parameters are identified and calculated synchronously byusing MDE/QDE. MDE is a modified di?erential algorithm for dealingwith the numerical part of the encoding scheme, and QDE is used fordealing with the binary part of that.4. For constrained optimization problem with the local minima, the tabu-based di?erential evolution is proposed. Tabu search compromises toaccept non-improved solutions, and can escape from the local minima. It is introduced the moving strategy in the neighboring area, tabu list,comprising list and aspiration criterion of tabu search to di?erential evo-lution. Moreover, the step length of tabu search adjusted dynamically isused in order to realize the self-adaptive search process. The experimen-tal results verifies that the proposed algorithm can enhance the ability ofescaping from the local minima.