Study On Nonlinear Neural Networks Based On Exact Penalty Function With Two-parameter

The NP-hard problems and the global optimization problems are two categories of the most challenged subjects in mathematical programming. Recently, some heuristic algorithms, such as simulated annealing (SA), genetic algorithms (GA) and artificial neural networks (ANN), are employed to these difficult problems and many of promising result are obtained. However, the amount of calculation, the speed of the convergence and the sensitivity of parameters of these algorithms are less encouraged. In the present paper, we investigated how to apply the artificial neural networks and genetic algorithm to the optimization problems with exact penalty function with two-parameterWe introduced the research condition of the penalty function, neural network and the genetic algorithm. Pass to the discussion of the exact penalty function with two parameter, make use of the property of the exact penalty function with two parameter, construct a kind of nonlinear nerve network, discussed the stability of neural network, submit an algorithm to get solution of this kind of neural networks, and give some examples. Finally, we investigated how to apply the genetic algorithm to the optimization problems with exact penalty function with two-parameter. We submit a genetic algorithm of mix for solving neural network, and give some numerical computations. The computation result expresses that it is availability. Compare to the pure genetic algorithm to solve the TSP, the new mixed method is more excellent in the astringency of solve.Main fruit is: building a model of nonlinear neural network with exact penalty function with two parameters, bring forward a simulate algorithm to get the solution, contemporary put forward a genetic algorithm of mix for solving neural network.This text has the important theory meaning for the theory of nonlinear network and solves the combinatorial optimization.