Incorporating elastic ring into Potts neural network
Jiann-Ming Wu, Cheng‐Yuan Liou · 2005
Devises a mechanism to include the elastic ring in the Potts neural encoding approach to solve the travelling salesman problem (TSP) with better performance. The obtained network is then called the elastic Potts neural network (EPNN) which has much looser coupled interconnections. Its function is close to the elastic ring method with different controlling mechanism for the ring path. The EPNN requires much less computation in finding a solution including the geometrical abstraction of this elastic ring for the TSP. Based on averages of simulations, the authors find that the performance of EPNN is much better than that of the simulated annealing method. The EPNN shows high stability in its solutions. The problem size can be extended easily to a 500-city TSP on a workstation.