A simple and high performance neural network for quadratic programming problems

In this paper, a neural network for quadratic programming problems is simplified. The simplicity is necessary for the high accuracy of solutions and low cost of implementation. The proposed network is proved to be an extension of Newton's optimal descent flow about constraints problems and is globally convergent. The network dynamic behaviors are also discussed and these can get the feasible solution more easily. The simulations demonstrate the reasonability of the theory and advantages of the network.