Determination of parameters in a Hopfield/Tank computational network

Hedge, Sweet, Levy
1988 IEEE International Conference on Neural Networks  
Neural-like networks which minimize a global energy function have been proposed for solving computationally intensive optimization problems. These networks have several parameters that need to be selected, and often, carefully tuned for a network to produce a sensible computation. We examine the Traveling Salesperson Problem (TSP) as a representative NP-complete optimization problem and present a cookbook approach to setting these parameters. Importantly, there appears to be a linear
more » ... p between two of these parameters. The relationship between these two parameters and the problem size leads to a simple understanding of why these networks are less and less useful for the TSP computation as the number of cities increases.
doi:10.1109/icnn.1988.23941 dblp:conf/icnn/HegdeSL88 fatcat:ond26qb23zhmhpzukjcia2bsiy