Intelligent Computing Techniques Via Small-world Network Model And Immune Clonal Optimization

Against the backgroud of information science, this dissertation focuses on the organic integrations of small-world network model and immune clonal optimization, and combines related contents of multi-discipline different subjects for various problems. Here on the one hand the complex network is the tool in study of intelligent optimizational algorithm, on the other hand its own is our research object.On the basis of information transmission dynamic in small-world network, two models are proposed by combining immune computing, which named as Immune Optimization Model of Network Feature, IOM-NF, and Immune Optimization Model of Network Structure, IOM-NS. Based on the two models, multiple novel immune small-world optimization algorithms are proposed for different problems. The main innovation of this disseration can be summaized as follows.1) Based on the information transmission dynamic in small-world netwotk, Immune Optimization Model of Network Feature, IOM-NF, and Immune Optimization Model of Network Structure, IOM-NS have been constructed. IOM-NF model concerned the property of network topology, and the operators for new algorithms have the network featues. IOM-NS model focus on the network’s topology structure, and the study object is the swarm algorithm. These two models established sound theoretical foundation of our researches.2) Based on the IOM-NF model, we propose relevant algorithms for unconstrained function optimization, satisfiability problems and linear system approximation problems, and verify their effectiveness via exprimental study. For unconstrained funcion optimization, the Immune optimization algorithm based on social network searching model, SNSIA, is proposed. Simulation results via three group 28 benchmark functions show the good performance. For satisfiability problems, we introduce the quantum computing, and combine the synergy strategy in biophysics, then design Quantum inspired Immune Coevolutionary algorithm with Adaptive Small-world model named QICAS. The uniform random 3-SAT problems are used to test the proposed algorithm’s performance. All results get high successes ratio and small computational complexity.For optimal approximation of linear systems, we proposed He Xie Evolutionary Immune information Network optimizer, abbreviated as HEIN. In HEIN, the population’s diversity remains well via the effective combination of ‘He’ Rules and ‘Xie’ Rules. The experiments on both stable and unstable linear system are obtained satisfactory results.3) Based on the IOM-NS model, the typical example of swarm algorithm called Particle Swarm Optimization(PSO) is our study object. We propose Dynamic Network-structure Immune PSO with Newman-Watts small-world topology, called DNIPSO for constrained benchmark optimization. Initially the topology of the population is the k neighbor coupled network. Then the Newman-Watts small world topology is formed gradually and the swarm evolves simultaneously. The optimization process contains the population structure dynamics and particle immune learning two parts which mutually promoted effectively in whole population. When the algorithm evolved to final stage, the topology of the particle swarm is the all connected network, which marked the whole evolution of population’s structure finished. Furthermore, the immune operator for pbest particle which is based on the clonal selection theory achieves a trade-off between exploration and exploitation abilities. Numerical experiments results on constrained benchmark functions show it is effective and robust.4) For society network clustering, the immune memetic algorithm and the strong-weak ties theory is combined to form the Immune Memetic Relationship Learning Algorithm,which is named as IMRLA. Here we take the complex network as our research object, and explore the community structure via the immune optimization of modularity function. The performance of our approach is tested on three wellknown society network. The results demonstrate that IMRLA get accurate division of society communities.5) For image segmentation, the immune small-world optimization algorithm is combined with threshold evaluation method, and the indivuals co-evolutionary strategy is designed for the image segmentation. The image objective and background region could be divided correctly through the best threshold getting from optimizing the information entropy function. These studies not only verify the performance of immune algorithm based on network property, but give the practical application of multi-disciplinary integration a new frontier.