神经网络模型求解约束优化问题的研究

Study on Neural Network Applied to Constrained Optimization

依据罚函数及动态处罚法设计增广Lagrange乘子函数,获得新的神经网络模型解决约束优化问题。理论论证了该模型的稳定性以及在一定条件下网络收敛的平衡点即为所解决的优化问题的最优解。事例仿真论证了该模型处理优化问题的有效性。

A novel neural network, of which is applied to universal nonlinear constrained optimization, is proposed based dynamical penalty function method and general Lagrange multiply one. Theoretically, its stability is demonstrated, and it converges to the optimal solutions of solved optimization problems under certain conditions. The simulation shows its availability.