Design of effective neural network ensembles for image classification purposes

In the field of pattern recognition, the combination of an ensemble of neur al networks has been proposed as an approach to the development of high per formance image classification systems. However, previous work clearly showe d that such image classification systems are effective only if the neural n etworks forming them make different errors. Therefore, the fundamental need for methods aimed to design ensembles of 'error-independent' networks is c urrently acknowledged. In this paper, an approach to the automatic design o f effective neural network ensembles is proposed. Given an initial large se t of neural networks, our approach is aimed to select the subset formed by the most error-independent nets. Reported results on the classification of multisensor remote-sensing images show that this approach allows one to des ign effective neural network ensembles. (C) 2001 Elsevier Science B.V. All rights reserved.