Attention improves the recognition reliability of backpropagation network

Authors:
Zbigniew Mikrut;Agata Piaskowska
Affiliations:
Institute of Automatics, AGH University of Science and Technology, Krakow, Poland;Institute of Automatics, AGH University of Science and Technology, Krakow, Poland
Venue:
ICAISC'06 Proceedings of the 8th international conference on Artificial Intelligence and Soft Computing
Year:
2006

Citing 2
Cited 0

A Selective Attention-Based Method for Visual Pattern Recognition with Application to Handwritten Digit Recognition and Face Recognition

IEEE Transactions on Pattern Analysis and Machine Intelligence
Preintegration lateral inhibition enhances unsupervised learning

Neural Computation

Quantified Score

Hi-index	0.00

Visualization

Abstract

In the paper a method is presented for improving the recognition reliability of backpropagation-type networks, based on the attention shifting technique. The mechanism is turned on in cases when the reliability of the network's answer is low. The signals reaching the hidden layer are used for selection of image areas which are the most ”doubtful” in the process of recognition by the network. Three methods have been proposed for appending the input vector after shifting the area where the attention is focused. The methods have been tested in the problem of hand-written digits recognition. Noticeable improvement of the recognition reliability has been obtained.