Impact of energy function on a neural network model for optimization problems

A neural network model for the solution of optimization problems is studied in this paper. Six different variations of the energy function that governs this network are presented and extensively evaluated. In order to evaluate these energy functions the traveling salesman problem has been used. The evaluation consists of measuring (over a large number of simulations) the following parameters: number of times that the network converges to a valid solution, quality of generated solution, and the number of network update cycles to reach a solution.<
>