The Research Of Two-stage Genetic Algorithm Optimization For Solving Dynamic Causal Model

Color and shape are the two basic features of images. Studies have shown that the brain handled color and shape separately. In order to identify images, the brain must bind color and shape feature eventually. Research on brain cognition has been a hot spot. Recently many research on human visual system has been done, which deepened the understanding of people about vision. However, storage bundles of image low-level features were neglected in the study of visual. This article want to study the process of binding shape and color by brain and construct dynamic causal models. In order to promote computer visual simulation, this article research on formation of Visual based on image low-level features levels and explore the human visual system.About Bundling for color and shape and constructing the DCM based on brain cognition network, the main work and innovation of this paper are as follows:1) introduce the theory about dynamic causal model and construct models which base on fMRI data of binding color and shape,using dynamic causal modeling methods in the past. Find the best model of binding color and shape by using bayesian model selection. And explain defects and deficiencies of this modeling.2) In view of flaws and disadvantages of dynamic causal modeling, this paper will apply classic genetic algorithm to optimize the construction of dynamic causal models and demonstrate that it is applicable. However, efficiencies of optimization is low in the stage when we apply the classic genetic algorithm into find the best DCM.3) This papers firstly apply the different two genetic algorithms on a common set of and compare the results by designing experiments, which proved two-stage genetic algorithm have higher efficiency of optimization, more stable and more excellent results, compared to the classic DCM optimization. Lastly, this paper construct dynamic causal models, applying two algorithms on the dataset which bounding shape and color. And we construct a better model for bounding color and shape.The two-stage genetic algorithm of this paper could work well with finding the best dynamic causal model, which have humanization, don’t require prior knowledge of guidelines, easy to operate and so on, compare to previous method of constructing dynamic causal models. The most import is that the two-stage genetic algorithm could be able to find the model which is more excellent than ever.