Gradient projection network: Analog solver for linearly constrained nonlinear programming

Authors:
Kiichi Urahama
Affiliations:
Department of Visual Communication Design, Kyushu Institute of Design, Shiobaru, Fukuoka, 815 Japan
Venue:
Neural Computation
Year:
1996

Citing 2
Cited 7

A new polynomial-time algorithm for linear programming

Combinatorica
Review:

Neural Computation

A deterministic annealing algorithm for approximating a solution of the max-bisection problem

Neural Networks
A Lagrange multiplier and Hopfield-type barrier function method for the traveling salesman problem

Neural Computation
On Convergence Conditions of an Extended Projection Neural Network

Neural Computation
A cooperative recurrent neural network for solving l1 estimation problems with general linear constraints

Neural Computation
A deterministic annealing algorithm for approximating a solution of the min-bisection problem

Neural Networks
A deterministic annealing algorithm for the minimum concave cost network flow problem

Neural Networks
A deterministic annealing algorithm for approximating a solution of the linearly constrained nonconvex quadratic minimization problem

Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

An analog approach is presented for solving nonlinear programming problems with linear constraint conditions. The present method is based on transformation of variables with exponential functions, which enables every trajectory to pass through an interior of feasible regions along a gradient direction projected onto the feasible space. Convergence of its trajectory to the solution of optimization problems is guaranteed and it is shown that the present scheme is an extension of the affine scaling method for linear programming to nonlinear programs under a slight modification of Riemannian metric. An analog electronic circuit is also presented that implements the proposed scheme in real time.