Information Sharing Mechanisms In Quantum Inspired Evolutionary Algorithms

Quantum-Inspired Evolutionary Algorithm (QIEA) is a new type of Estimation of Distribution Algorithms (EDAs), which utilizes multiple co-evolved quantum probability models to guide the search of the algorithm. Most of previous EDAs have shown pretty good convergence speed, but the exploration ability of them is not satisfying when applied on complex multimodal optimization problems. Base on the multi-model structure of QIEA, this dissertation studies the information sharing mechanism in QIEA, the issue of balancing the convergence and exploration ability of QIEA, and then studies new and more efficient QIEAs. The topics of this dissertation are intensively concerned in the field of evolutionary computation, therefore have important research value and promising application perspective.The major works and contributions of this dissertation include:1) Several information sharing mechanisms of QIEAs are experimentally analyzed based on large scale 0-1 knapsack problems and NK fitness landscape model. The results show that the information sharing has important influence on the performance of QIEAs. The ability of finding optimal solutions of QIEAs with information sharing is obviously better than that without information sharing. Stochastic and comprehensive information sharing is better for QIEA to deal with complex large scale optimization problems than information sharing based on the fixed neighborhood structure.2) A new QIEA (CLQEA) based on comprehensive learning is proposed, in which a comprehensive, bit-bit information sharing mechanism is implemented based on the idea of comprehensive learning. Experimental results show that CLQEA has better performance on large scale complicated optimization problems, while also has good exploration ability.3) A new QIEA (QBO) based on the biogeography model is proposed, in which comprehensive information sharing mechanism among quantum individuals is implemented in a more natural way via simulating the migration behavior among multiple habitats. Experimental results show that QBO can share good knowledge among models in a more efficient way, which helps QBO to find better solutions with relatively lower computational cost.The work of this dissertation will be helpful to reveal the principle of QIEAs in more depth, and therefore be helpful to design more effective QIEAs. Besides, the new QIEAs proposed are applicable to many complex and large scale optimization problems in real engineering and scientific applications.