A novel multicriteria optimization algorithm for the structure determination of multilayer feedforward neural networks

We propose in this paper a novel prescriptive solution to decide the optimum number of neurons in the hidden-layer of multilayer feedforward neural networks. Our approach uses the unconstrained mixed integer nonlinear multicriteria optimization technique. We validate the algorithm using numerical examples. We extend the above results using fuzzy reasoning and constrained optimization techniques to solve the cross-validation problem in a more effective way than the traditional back propagation algorithm. The main features of our approach are that it is a formal method and it draws results from many fields to combine them for the solution of this NP-complete problem.