Globally Convergent Modification of the Quickprop Method

Authors:
Michael N. Vrahatis;George D. Magoulas;Vassilis P. Plagianakos
Affiliations:
Department of Mathematics, University of Patras, GR-261.10 Patras, Greece. E-mail: vrahatis@math.upatras.gr;Department of Informatics, University of Athens, GR-157.84 Athens, Greece.;University of Patras Artificial Intelligence Research Center–UPAIRC.
Venue:
Neural Processing Letters
Year:
2000

Citing 10
Cited 2

A view of unconstrained optimization

Optimization
Learning internal representations by error propagation

Parallel distributed processing: explorations in the microstructure of cognition, vol. 1
Rescaling of variables in back propagation learning

Neural Networks
Speed up learning and network optimization with extended back propagation

Neural Networks
Original Contribution: An analysis of premature saturation in back propagation learning

Neural Networks
Effective backpropagation training with variable stepsize

Neural Networks
Optimization: algorithms and consistent approximations

Optimization: algorithms and consistent approximations
Improving the convergence of the backpropagation algorithm using learning rate adaptation methods

Neural Computation
Iterative solution of nonlinear equations in several variables

Iterative solution of nonlinear equations in several variables
Numerical Methods for Unconstrained Optimization and Nonlinear Equations (Classics in Applied Mathematics, 16)

Numerical Methods for Unconstrained Optimization and Nonlinear Equations (Classics in Applied Mathematics, 16)

Self-scaled conjugate gradient training algorithms

Neurocomputing
Determining the number of real roots of polynomials through neural networks

Computers & Mathematics with Applications

Quantified Score

Hi-index	0.00

Visualization

Abstract

A mathematical framework for the convergence analysis of the well-known Quickprop method is described. Furthermore, we propose a modification of this method that exhibits improved convergence speed and stability, and, at the same time, alleviates the use of heuristic learning parameters. Simulations are conducted to compare and evaluate the performance of the new modified Quickprop algorithm with various popular training algorithms. The results of the experiments indicate that the increased convergence rates achieved by the proposed algorithm, affect by no means its generalization capability and stability.