Abstract: A kind of neural network model for nonlinear programming problems on a closed convex set is presented in this paper. Theoretic analysis and simulation results on the computershow the neural network is globally convergent exponentially to the exact solutions of theprogramming problems under some appropriate assumptions. The optimization method employed by theneural network falls into the extended gradient method. The model is simpler than the existing neural network models even when it is for positive definite quadratic programming problems.