Fast-learning variations on back propagation: an empirical study.

By Scott E. Fahlman. 1989

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Improved spiking neural networks for EEG classification and epilepsy and seizure detection (2007, unknown): “For QuickProp, the heuristic proposed by Fahlman [43] for the traditional ANN, that is, deﬁning a maximum growth factor ( µ ) as the upper limit for η Q is used.”
Evolving artificial neural networks (1999, unknown): “Evolutionary training can be slow for some problems in comparison with fast variants of BP [131] and conjugate gradient algorithms [19], [132].”
Neural networks for the prediction and forecasting of water resource variables: a review of modelling issues and applications (2000, methodology): “The Quickprop algorithm was also used by Karunanithi et al. (1994); Muttiah et al. (1997).”
The Recurrent Cascade-Correlation Architecture (1990, methodology): “This single layer of connections is trained (using the Quickprop algorithm [Fahlman, 1988]) to minimize the error.”
The Cascade-Correlation Learning Architecture (1989, unknown): “---o--------~&_-------mr_--------------~----~--~~ Inpu~ O--------~~----~H-------------~.---~~--~ o--------~~------~--------------------------~”
The Importance of Neurophysiological Constraints for Modelling the Emergence of Modularity (methodology): “The intermediate value of µ ~ 0.1 is equivalent to using SSE with a Sigmoid Prime Offset, which was suggested by Fahlman (1988) as a way of resolving some of the learning difficulties inherent in using pure SSE.”
Analysis of Dynamical Recognizers (1997, methodology): “, 1986), with a modi cation similar to Quickprop (Fahlman, 1989)1 and a learning rate of 0:03.”
Empirical evaluation of the improved Rprop learning algorithms (2003, methodology): “Fahlman's Quickprop (cf. [2] [15]), the BFGS (cf. [14]) algorithm, a quasi-Newton method, which iteratively estimates the inverse of the Hessian, and the conjugate gradient method (cf. [14] [20]).”
Scaling Large Learning Problems with Hard Parallel Mixtures (2002, background): “(3) Train each local gater P (X |E = i) separately over Di.”
Numerical implementation of a neural network based material model in finite element analysis (2004, methodology): “In updating connection weights of the neural networks, many variations of the back-propagation algorithm have been proposed [ 22–25 ] .”
Nonlinear principal component analysis using autoassociative neural networks (1991, background): “This has 3een a popular test problem for network training methodologits (Rumelhart et al., 1986; Fahlman, 1989).”
SEMINT: A tool for identifying attribute correspondences in heterogeneous databases using neural networks (2000, methodology): “We have implemented two versions of back-propagation algorithms, standard [27,28] and quick-propagation [6].”
Learning Rate Adaptation in Stochastic Gradient Descent (2001, background): “Methods suited to on–line learning are those that can handle nonstationary (time–varying) data, while at the same time, require relatively little additional memory and computation, in order to process one additional example.”
Lamarckian Evolution and the Baldwin Effect in Evolutionary Neural Networks (2006, methodology): “G-Prop (a genetic evolution of BP trained MLP), used in this paper to tune learning parameters and to set the initial weights and hidden layer size of a MLP, searches for the optimal set of weights, the optimal topology and learning parameters, using an EA and Quick-Propagation [18] (QP).”