平滑的
人工神经网络
惯性参考系
计算机科学
数学优化
最优化问题
投影(关系代数)
光学(聚焦)
功能(生物学)
数学
最优控制
函数逼近
控制理论(社会学)
点(几何)
随机神经网络
投影法
班级(哲学)
算法
趋同(经济学)
理想(伦理)
多目标优化
全局优化
作者
Jie Shen,Lu Yuan,Yuzhu Tian,Song Yu-Hui
标识
DOI:10.61208/pjo-2025-050
摘要
(Communicated by Jie Sun) With the increasing complexity of the problems in practical fields, traditional numerical algorithms are not very ideal for solving non-smooth non-Lipschitz optimization problems. We focus on a broader class of non-smooth non-Lipschitz optimization problems, in which the objective function is the sum of a non-smooth non-convex function and a non-Lipschitz function. Based on the smoothing approximation technique and projection neural network, we propose a smooth inertial projection neural network (SIPNN) by introducing inertial terms into the neural network. Under certain conditions we show that the solution of SIPNN is convergent to the stationary point of non-Lipschitz optimization problems. Compared with other methods employing differential inclusions for solving non-smooth problems, the smoothing approximation technique can effectively overcome the difficulty of choosing subdifferentials, and the introduction of inertial terms can control variables which allows us to explore globally the optimal solution of the original optimization problems. The simulation results are presented to illustrate the performance and effectiveness of the proposed method.
科研通智能强力驱动
Strongly Powered by AbleSci AI