In the last few years, several new results on Self-Organizing Map algorithms applied to the Euclidean Travelling Salesman Problem (ETSP) have emerged. All of them have attempted to find quasi-optimal solutions for this NP-complete combinatorialproblem based on the functional role of the learning process mapping understood to occur in the brain.This paper brings an innovative computational investigation based on a new model of lateral interactions between neurons that can be thought of as introducing co-operation between them. Such interactions are now being observed in biological neural nets.Its application to ETSP obtained results never before achieved. Simulations using a sequential machine for well-known and difficult TSP library instances as well as for problems with over two thousands cities will be discussed in this paper.
展开▼