A novel digital neural network for the travelling salesman problem

In this paper, we present a novel neural network solution to the Travelling Salesperson´s Problem. In comparison to the Hopfield Neural Network for the task, the proposed network requires N neurons and O(N²) weights to solve a N-city instance. The proposed network is entirely digital, and has been mapped onto a Field Programmable Gate Array. Initial results are encouraging and the FPGA takes under 20 microseconds to solve the task for chosen examples. The re-programmable nature of the FPGA makes the proposed approach a very low cost, reusable hardware approach.