A New Optimization Method Based on Dynamic Neural Networks for Solving Non-convex Quadratic Constrained Optimization Problems

This paper presents a capable recurrent neural network, the so-called µRNN for solving a class of non-convex quadratic programming problems‎. ‎Based on the optimality conditions we construct a new recurrent neural network (µRNN)‎, ‎which has a simple structure and its capability is preserved‎. ‎The proposed neural network model is stable in the sense of Lyapunov and converges to the exact optimal solution of the original problem‎. ‎In a particular case‎, ‎the optimality conditions of the problem become necessary and sufficient‎. ‎Numerical experiments and comparisons with some existing algorithms are presented to illustrate the theoretical results and show the efficiency of the proposed network.