The authors propose the architecture of a hybrid Neo-ART/EBP (adaptive resonance theory/error-back-propagation) neural network and describe the results that may be achieved for digit recognition applications. Joining together a simplified input ART layer and an output EBP network makes it possible to reduce the global number of hidden nodes/interconnections and to speed up the convergence time during the training phase. Different strategies are exploited during the learning step to achieve lower total error and faster convergence time. Moreover, in the pattern space, both circular and elliptical regions are investigated, and their influence is discussed.
A Hybrid NeoART/EBP Architecture for hand-written digit recognition
V Rampa;
1992
Abstract
The authors propose the architecture of a hybrid Neo-ART/EBP (adaptive resonance theory/error-back-propagation) neural network and describe the results that may be achieved for digit recognition applications. Joining together a simplified input ART layer and an output EBP network makes it possible to reduce the global number of hidden nodes/interconnections and to speed up the convergence time during the training phase. Different strategies are exploited during the learning step to achieve lower total error and faster convergence time. Moreover, in the pattern space, both circular and elliptical regions are investigated, and their influence is discussed.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
