حل مساله تخصيص تعميم يافته به روش شبكه عصبي

Solving Generalized Assignment Problem with Neural Network

In this paper we apply neural network method for solving generalized assignment problem. This problem is the generalization of the well-known assignment problem that can be formulated as a zero-one integer programming problem. For solving this problem applying neural network method, first we transform it to a nonlinear programming problem, then we use two types of neural network structure one based on penalty function and the other augmented lagrangean multiplier method and compare them with each other. It will be noticed that neural network based on penalty function method is not appropriate to combinatorial optimization problems because it reaches an infeasible solution or gets stuck in a nonoptimal feasible solution while neural network based on augmented lagrangean multiplier method can alleviate this problem to some extent.