Trust region method in neural network

Trust region method in neural network

0.00 Avg rating0 Votes
Article ID: iaor19981850
Country: China
Volume: 13
Issue: 4
Start Page Number: 342
End Page Number: 352
Publication Date: Oct 1997
Journal: Acta Mathematicae Applicanda Sinica
Authors:
Keywords: programming: quadratic, programming: nonlinear
Abstract:

A Hopfield-type neural network with adaptively changing synaptic weights and activation function parameters is presented to solve unconstrained nonlinear programming problems. The network performance is similar to that of the trust region method in the mathematical programming literature. There is a sub-network to solve quadratic programming problems with simple upper and lower bounds. By sequentially activating the sub-network under the control of an external computer or a special analog or digital processor that adjusts the weights and parameters, the network solves a sequence of unconstrained nonlinear programming problems. Convergence proof and numerical results are given.

Reviews

Required fields are marked *. Your email address will not be published.