An active learning paradigm based on a priori data reduction and organization

• Published in: Expert systems with applications Vol. 41; no. 14; pp. 6086 - 6097
• Main Authors: Saito, Priscila T.M., de Rezende, Pedro J., Falcão, Alexandre X., Suzuki, Celso T.N., Gomes, Jancarlo F.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier Ltd 15.10.2014; Elsevier
• Subjects: Active learning; Applied sciences; Artificial intelligence; Classification; Computation; Computer science; control theory; systems; Data mining; Data processing. List processing. Character string processing; Data reduction; Exact sciences and technology; Expert systems; Image annotation; Information systems. Data bases; Iterative methods; Learning; Machine learning; Memory organisation. Data processing; Organizations; Pattern recognition; Pattern recognition. Digital image processing. Computational geometry; Software; Strategy; Pattern recognition; Data mining; Active learning; Image annotation; Machine learning; Image processing; High precision; Very large databases; Active system; Efficiency; User interface; Classification; Learning algorithm; Small medium sized firm; Data analysis; Process selection; Data reduction; Interactive system; Dimension reduction; Experimental result; Supervised learning; Learning (artificial intelligence); Small sample; Artificial intelligence; Indexing
• DOI: 10.1016/j.eswa.2014.04.007
• ISSN: 0957-4174

•A novel active learning paradigm, called DROP, based on a priori data reduction and organization.•DROP does not require classification and reorganization of all non-annotated samples in the dataset at each iteration.•The proposed paradigm allows to achieve high accuracy quickly with minimum user interaction.•Results are shown with different clustering and classification strategies, and on a variety of real-world datasets. In the past few years, active learning has been reasonably successful and it has drawn a lot of attention. However, recent active learning methods have focused on strategies in which a large unlabeled dataset has to be reprocessed at each learning iteration. As the datasets grow, these strategies become inefficient or even a tremendous computational challenge. In order to address these issues, we propose an effective and efficient active learning paradigm which attains a significant reduction in the size of the learning set by applying an a priori process of identification and organization of a small relevant subset. Furthermore, the concomitant classification and selection processes enable the classification of a very small number of samples, while selecting the informative ones. Experimental results showed that the proposed paradigm allows to achieve high accuracy quickly with minimum user interaction, further improving its efficiency.