Recurrent neural network for complex-variable pseudoconvex optimization with equality constraints

摘要

In this paper, for solving a complex-variable pseudoconvex optimization with equality constraints, a one-layer recurrent neural network is proposed. From any initial point, the state of the presented neural network is proved to converge exponentially to the feasible region of the pseudoconvex optimization problem. When the initial point belongs to the feasible region, the state solution will converge to an optimal solution of the complex-variable pseudoconvex optimization ultimately. Compared with known neural networks for complex-variable pseudoconvex optimization, the model of the proposed neural network is less complex. Finally, numerical examples are provided to substantiate the feasibility of the proposed neural network.