A Log-Sigmoid Lagrangian Neural Network for Solving Nonlinear Programming

A similar neural network as Zhang and Constantinides [9] is proposed in this paper for solving nonlinear programming with equality and inequality constraints. We overcome the condition of positive definiteness of Lagrangian Hessian by introducing the Log-Sigmoid (LS) function. Thus the proposed network is simpler than the augmented lagrangian neural network inform and have weaker condition than lagrangian neural network in [9].