A Novel Fuzzy Approach to Hopfield Coefficients Determination
Salvatore Cavalieri, Marco Antonio Russo · 2024
The Hopfield-type neural model is a suitable tool for the solution of optimization problems featuring NP-hard computational complexity. The solution of such problems using a Hopfield model requires determination of the values of a certain number of coefficients linked to the surrounding conditions of the optimization problem itself. It is quite difficult to determine these values, because a heuristic search is necessary. This is not only time-consuming, but may lead to the determination of a set of coefficients which provide neural solutions that are far from optimal, or even non-valid ones. So far there have been no reported research in literature offering a general method for the search for coefficients which will guarantee optimal or close to optimal solutions. This chapter proposes a method which allows automatic determination of Hopfield coefficients. The results obtainable can be user-defined. It is, in fact, possible to specify whether the user wishes high quality solutions or prefers greater robustness, i.e., a high percentage of valid solutions.