A feedback neural network for solving convex constraint optimization problems

In this paper, a feedback neural network model is presented by two steps. Firstly, a convex sub-optimization problem with bound constraints is established by introducing an energy function and the neural subnetwork for solving the sub-optimization problem is constructed based on the projection method. Secondly, a feedback neural network is proposed by using the subnetwork and can converge to an exact optimal solution of primal optimization problem. The distinguishing features of the proposed feedback network are no Lagrange multipliers, no dual variables, and no penalty parameters. It has the least number of state variables, simple structure, and is suitable for hardware implementation. Two simulation examples are provided to show the feasibility and efficiency of proposed method in the paper.