A Survey of Neurodynamic Optimization

Abstract

The last four decades have witnessed the birth and growth of neurodynamic optimization with numerous recurrent neural networks developed for solving various constrained optimization problems. Numerous results on neurodynamic optimization are reported in the literature,. In view of the diverse nature of the publications, this survey provides an updated overview of neurodynamic optimization to summarize the state-of-the-art results in terms of model structure, convergence property, and solvability scopes. It starts with an introduction and preliminaries, followed by categorizing many representative neural network models for constrained optimization, such as linear and quadratic programming, smooth and nonsmooth nonlinear programming, minimax optimization, distributed optimization, generalized-convex optimization, and global and mixed-integer optimization. In addition, it also delineates some perspective research topics for further investigations.