Chaotic Neural Network-Based Approach to the Shortest Path Problem
Xiu Wang · Systems Enging-theory Methodology Application · 2001
The shortest path problem is one of classical combinatorial optimization problems arising in numerous planning and designing contexts. However, the conventional Hopfied neural networks (HNN) for the shortest path problem is easy to get stuck into local minima. In this paper, by introducing a time variant parameter to control the chaos, the chaotic neural network (CNN) becomes a transiently chaotic neural network (TCNN), which can converge to a stable global optimal solution. Then, neural netowrk and computational energy for solving the shortest path problem is presented. Numerical simulations show that TCNN in solving the shortest path problem always converges to the globally optimal solution and has higher efficiency of searching than HNN.