Title: About retraining rule in multi-expert intelligent system for semi-supervised learning using SVM classifiers

Authors: D. Barbuzzi; G. Pirlo; D. Impedovo

Addresses: Department of Computer Science, University of Bari 'Aldo Moro', Bari, Italy ' Department of Computer Science, University of Bari 'Aldo Moro', Bari, Italy ' Department of Electrical and Electronic Engineering, Polytechnic of Bari, Bari, Italy

Abstract: This paper proposes three methods in order to retrain classifiers in a multi-expert scenario, when new (unknown) data are available. In fact, when a multi-expert system is adopted, the collective behaviour of classifiers can be used for both recognition aims and selection of the most profitable samples for system retraining. More specifically a misclassified sample for a particular expert can be used to update the expert itself if the collective behaviour of the multi-expert system allows the classification of the sample with high confidence. In addition, this paper provides a comparison between the new approach and those available in the literature for semi-supervised learning using the SVM classifier by taking into account four different combination techniques at abstract and measurement levels. The experimental results, which have been obtained using the handwritten digits of the CEDAR database, demonstrate the effectiveness of the proposed approach.

Keywords: feedback-based strategies; semi-supervised learning; intelligent multi-expert systems; expert systems; retraining rule; SVM classifiers; support vector machines; classifier retraining.

DOI: 10.1504/IJSISE.2014.066607

International Journal of Signal and Imaging Systems Engineering, 2014 Vol.7 No.4, pp.245 - 251

Received: 19 Apr 2013
Accepted: 30 Sep 2013
Published online: 29 Dec 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article