Research on Optimization of Convolutional Neural Networks based on Variational Inequalities: From the Perspective of Topological Structure

Abstract

Based on the sufficient and necessary conditions of the solution, a neural network model for generalized variational inequality problems is proposed. Through the construction function, the new model is proved to be stable under appropriate conditions, and the global convergence and exponential convergence are consistent with the original problem. Numerical experiments show that the solution. The neural network model is effective and feasible. Starting from the nature of the variational inequality, a recurrent neural network is constructed to solve this type of optimization problem, and from the KKT condition of the optimization problem, a recurrent neural network is constructed to solve this type of optimization problem.