Abstract: A neural network model is constructed, and the stability of the model is proved to be an important problem in solving nonlinear optimization. Matrix variable neural network model is an extension of vector neural network. A large number of researchers have proved that the former has more advantages in computational speed and application. A new matrix projection neural network is proposed for a class of nonlinear programming with mixed constraints, and the global stability of the model is proved. The simulation experiments further verify the conclusion.

Key words: matrix projection neural network, nonlinear optimization, mixed constraints, global convergence, speed