Another simple recurrent neural network for quadratic and linear programming

Abstract

A new recurrent neural network is proposed for solving quadratic and linear programming problems, which is derived from two salient existing neural networks. One of the predecessors has lower structural complexity but were not shown to be capable of solving degenerate QP problems including LP problems while the other does not have this limitation but has higer structural complexity. The proposed model inherits the merits of both models and thus serves as a competitive alternative for solving QP and LP problems. Numerical simulations are provided to demonstrate the performance of the model and validate the theoretical results. © 2009 Springer Berlin Heidelberg.