The Simulation And Evolution Research Of The Visual Cortex Synaptic Connection Networks

The brain is one of the most efficient, most complex and stablest systems. Complex networks have become new methods to study the brain neurons connections and brain region connections. Existing research has shown that the brain networks have small world properties, that is to say they have shorter average path length and higher clustering coefficient simultaneously, corresponding to function segregation and function integration respectively, the two fundamental organizational principles of information processing brain. The visual cortex exists in multiple regions of the brain, which is a breakthrough to study the brain.The number of synaptic connections in the brain develops according to a model of increasing at first and then decreasing. The key point lies in the fact that synaptic pruning mechanism of the cerebral cortex may make the brain network efficient and steady. It also provides a new solution to create high efficiency and low energy consumption networks. In addition, most neurons connections in neural system are short, while there are also few long-distance connections. The internal mechanism that leads to the formation of synaptic connection mode is not clear. As for the minimization of both the average steps number of information transmission and the overall wiring cost, there may exist a trade-off between distance priority connection and large neurons priority connection in the visual network. The mathematical model is used in this thesis to verify the small world characteristics of visual cortex networks. Based on it, we study the visual cortex synaptic pruning mechanisms and synaptic connections mechanisms. The research contents are summarized as follows:Firstly, based on human brain voxels coordinates, the positions of the nodes are established under the strictly space limit condition of the brain. And then according to organizational structure features of cortex, the space distances between nodes are introduced in the visual cortex. The connection probability function between nodes is used as mathematical descriptions of network. The experiments of visual cortical networks are conducted within the local and global scope, which verifies the small world characteristics of visual cortex networks.Secondly, the network cutting edge evolution model is designed according to the synaptic pruning mechanisms in the cerebral cortex network. Experimental results show that the model can successfully dilute network under the condition of connectivity. The evoluted networks show the properties of hub nodes, small-world characteristics and high efficient-cost ratio and so on. It verifies the hypothesis that synaptic pruning mechanism may be crucial to make brain network efficient.Finally, to weigh the mechanism between the overall wiring cost and the information transmission efficiency in visual cortex, and combining with complex network dynamic evolutionary game theory, the network evolution game model based on large neurons priority and distance priority is established. Experimental results show that parts of close edges are replaced by long-distance edges in evoluted networks. The topology structure and statistical characteristics of networks are changed. When evolution games stabilize in the state that collaborators account for the most proportion, the network appeares relatively fewer long-range connections. This reflects the close connection priority mechanism. When evolution games stabilize in complete betrayal state, the networks appeare more long-distance connections, and appear hub nodes, which reflect the visual cortex of large neurons preferential attachment mechanism. This model can imitate a trade-off relationship between minimum wiring costs and minimum processing steps in the visual cortex.