Fly Immune Optimization Algorithms With Coevolution And Their Applications

Fruit flies are a class of arthropods with the ability of unique perception to the position and direction of food resource. Their metaphors and mechanisms of innate immune response, vision and smell give researchers many biologically theoretical foundations for promoting the development of intelligent science. Some high-efficient advanced intelligent methodologies can be developed by simulating their foraging behavior characteristics and also intrinsic immune mechanisms. This will become an important research direction in the branch of computational intelligence. Based on such considerations, this thesis studies several advanced fly optimization approaches according to simplified metaphors of vision, smell and immune response for fruit flies. Afterwards, such approaches are further investigated with respect to their computational complexity and experimentally comparative analysis. The achievements acquired are helpful for developing the branch of intelligent optimization, while being of important reference value for solving engineering optimization problems. The main work and achievements acquired are summarized below.A. A visual directional selective fly optimization approach is developed to solve non-constrained function optimization problems with higher dimensions. It bases on fly's visual perceptional performances to environments and olfactory sensitivity to food. Some biological behaviors of visual wide field neurons are simulated to design an individual update module which updates those inferior individuals in the current population. Some theoretical analyses indicate that the computational complexity of the approach is decided by population size and problem dimension. The comparative experiments have showed that one such approach can perform stable search and acquire high-quality solutions.B. In order to overcome the shortcoming of low efficiency for the above algorithm, an improved micro-population fly optimization algorithm is designed to solve complex non-constrained function optimization problems. It is with the merits of small population, few iterative numbers and low complexity, while being able to strengthen the diversity of population and enhance the quality of solution search by taking full advantage of coevolution between subpopulations in the process of evolution. The computational complexity analysis shows that the complexity of the approach is determined by population size and problem dimension. Experimental results have illustrated that one such improved approach can efficiently execute effective search while being a potential optimizer for solving optimization problems with higher dimensions.C. A fly immune optimization approach with coevolution is proposed based on the simplified process of immune response. In this approach, those high-quality individuals are produced through an inner loop, while an external loop is designed to accelerate the process of solution search. Suchtwo interactive loops can ensure that the approach keeps the diversity of population and strengthens the capability of global evolution. The numerical results have illustrated that one such approach can achieve stable search when addressing optimization problems with high dimensions.