Performance improvement of classifier fusion for batch samples based on upper integral. (March 2015)

• Record Type: Journal Article
• Authors: Feng, Hui-Min; Wang, Xi-Zhao
• Is Part Of: Neural networks. Volume 63(2015:Mar.)
• Page Start: 87
• Page End: 93
• Publication Date: 2015-03
• Subjects: Extreme learning machine -- Upper integral -- Fuzzy measure -- Fuzzy integral -- Multiple classifier fusion; Neural computers -- Periodicals; Neural networks (Computer science) -- Periodicals; Neural networks (Neurobiology) -- Periodicals; Nervous System -- Periodicals; Ordinateurs neuronaux -- Périodiques; Réseaux neuronaux (Informatique) -- Périodiques; Réseaux neuronaux (Neurobiologie) -- Périodiques; Neural computers; Neural networks (Computer science); Neural networks (Neurobiology); Periodicals; 006.32
• DOI: 10.1016/j.neunet.2014.11.004
• Languages: English
• ISSNs: 0893-6080

Abstract

Abstract: The generalization ability of ELM can be improved by fusing a number of individual ELMs. This paper proposes a new scheme of fusing ELMs based on upper integrals, which differs from all the existing fuzzy integral models of classifier fusion. The new scheme uses the upper integral to reasonably assign tested samples to different ELMs for maximizing the classification efficiency. By solving an optimization problem of upper integrals, we obtain the proportions of assigning samples to different ELMs and their combinations. The definition of upper integral guarantees such a conclusion that the classification accuracy of the fused ELM is not less than that of any individual ELM theoretically. Numerical simulations demonstrate that most existing fusion methodologies such as Bagging and Boosting can be improved by our upper integral model.