A single layer recurrent neural network for pseudoconvex optimization subject to quasiconvex constraints

This paper presents a single layer recurrent network for solving optimization problems with pseudoconvex objectives subject to quasiconvex constraints. The penalty method using a finite penalty parameter is applied for the design and analysis of the neural network. The lower bounder of the penalty parameter is given in order to guarantee the exact penalty property. It is rigorously proved that the neural network is globally convergent to the global optimal solution of the corresponding optimization problem. Simulation results are included to illustrate the performances of the proposed neural network.