Research On Shortest Path Model And Algorithm Of Fuzzy And Time-varying Neural Network

Dynamic network shortest path problem is one of the most important part of network optimization, the traditional algorithms such as Dijkstra?A * will bring errors in solving the shortest path problem of dynamic network, and intelligent algorithms such as genetic algorithm?ant colony optimization algorithm have disadvantage in iterations and efficiency.We try to design automatic wave neural network algorithm for solving the problem better,there are three parts as bellow.(1) Given the fuzzy expected shortest path problem definition in order to solve shortest path problem of fuzzy network, proposed PFNNSP algorithm based-on automatic wave, by the way stated the PFNNSP algorithm's process of running in simple example, the test in random generation fuzzy network shows that PFNNSP algorithm has better performance than Dijkstra algorithm in running time, and the test in international dataset shows that PFNNSP algorithm has better performance than Dijkstra algorithm and A* algorithm in numbers of iterations and rate of convergence.(2) Given the shortest path problem definition in order to solve shortest path problem of dynamic weight network, to deduce formal representation through the bus model( the value of edges' weight are assumed as exponential distribution).Proposed STDNN algorithm based-on automatic wave, take into account the complexity of algorithm designation, STDNN algorithm will transform uncertain edges' weight into certain weight by random simulation. The test in dataset shows that STDNN algorithm has better performance than DPA algorithm in running time for the dense dynamic weight network,and there are no differences for the two algorithms in solving static shortest path problem.(3) Given the definition of time-varying with time windows in order to study shortest path problem in time-varying network, and design the neuron with time windows according to features of time-varying network, next proposed TDNN algorithm based-on automatic wave, by the way give the theorem of optimal solution, automatic wave activations theorem, and analyzed TSNN's time and space complexity, the test on small scale dynamic dataset shows that TDNN algorithm has better performance than PCNN algorithm in running time, and the test on big scale static dataset shows that TDNN algorithm has better performance than Dijkstra algorithm in running time, at last, TDNN algorithm solve the dense network shortest path problem better because automatic wave carries more status information during run time.