Accumulative competition neural network for shortest path tree computation
Jiyang Dong, Wen-Jun Wang, Junying Zhang · 2004
Shortest path tree (SPT) computation is an important combinatorial optimization problem with numerous applications. A novel neural network model called accumulative competition neural network (ACNN) is proposed in this paper to compute the SPT in a given weighted graph. Comparing with the other neural network based search algorithms, the algorithm presented here features in much less number of neurons needed, much less iterations of the network needed, the simplicity of neuron model, the simplicity of the topology structure of the network and the global optimal solution result. Finally, examples for searching the shortest path tree in weighted graphs are given. All the results have shown the high performance of the ACNN in searching the SPT in weighted graph.