Digital.CSIC > Ciencia y Tecnologías Físicas > Instituto de Robótica e Informática Industrial (IRII) > (IRII) Artículos >




Open Access item A deterministic algorithm that emulates learning with random weights

Authors:Ruíz de Angulo García, Vicente
Torras Genís, Carme
Keywords:Noise, Noisy training, Regularization, Fault-tolerance, Optimization, Sensitivity, Adaptive systems, Automation
Issue Date:2002
Citation:Neurocomputing 48(1): 975-1002 (2002)
Abstract:The expectation of a function of random variables can be modeled as the value of the function in the mean value of the variables plus a penalty term. Here, this penalty term is calculated exactly, and the properties of different approximations are analyzed. Then, a deterministic algorithm for minimizing the expected error of a feedforward network of random weights is presented. Given a particular feedforward network architecture and a training set, this algorithm accurately finds the weight configuration that makes the network response most resistant to a class of weight perturbations. Finally, the study of the most stable configurations of a network unravels some undesirable properties of networks with asymmetric activation functions.
Publisher version (URL):http://dx.doi.org/10.1016/S0925-2312(01)00695-6
Appears in Collections:(IRII) Artículos

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.