Research On Method Of Modelling And Optimization In Net-Flow System

Net-flow systems play an important role in economic society. At present, there are many methods of modelling net-flow systems. However, most of these modelling methods only describe the static characteristics of system. In order to rightly analyze and design these systems, it is very necessary to build up appropriate model of system.This paper begins with the restrictions of the-existent net-flow modelling methods. Integrated with the characteristics of modern net-flow system and utilizing Francesco Luna's idea of pcrceptron and XOR function, we express human's consciousness with logistic function of perceptron, and the uncertain relations among nodes are mapped to XOR function. On the basis of these, a new modelling method of net-flow is offered.The characteristic of flow in complex system is expressed through dynamic process. The form of net-flow is the result of dynamic game among nodes. The choice of infuse nodes has certain selectivity and this kind of selectivity is analogous to effect of human's consciousness. The flowing process is processed in a certain space, and the environment, the formed structure and flowing process change with time. According these, the dynamic characteristic of net-flow is an important problem that should be firstly considered in modelling. The connecting states of nodes are uncertain and the linking relation between two adjacent nodes is unperceptible in advance, so this unperceptivity can be mapped to XORfunction; When forming the access of flow, not all nodes can become the node on the access. To a certain degree, the form of access embodies intelligence of human, this ability of choice can be mapped to logistic function. When estimate the connecting relation among nodes, we utilizethe matching relation between enlarged neural network and XOR function. This is a learning process and it is a dynamic process with ceaseless game.This paper provides an arithmetic of path optimization in net-flow system. After the connecting relations are received with the process of dynamic game, the optimistic path can be found by this method. This method of path optimization can directly gain access of net-flow via result of operation. It doesn't need reversibly operation to find optimistic path.In this paper, Swarm software is utilized to simulate the model built through above method. The continually optimized connecting relation is gained via perceptron and XOR function, then the optimal path graph is found. This paper analyzes the result of simulation by using Luna's benchmark table and indicates that the distribution of nodes with different perceptivity is a hiding order of different optimal path produced by system.