Ruta, D. and Gabrys, B., 2002. A theoretical analysis of the limits of majority voting errors for multiple classifier systems. Pattern Analysis and Applications, 5 (4), pp. 333-350.
Full text not available from this repository.
Official URL: http://www.springerlink.com/content/3w8ugduqtcb2ap...
A robust character of combining diverse classifiers using a majority voting has recently been illustrated in the pattern recognition literature. Furthermore, negatively correlated classifiers turned out to offer further improvement of the majority voting performance even comparing to the idealised model with independent classifiers. However, negatively correlated classifiers represent a very unlikely situation in real-world classification problems, and their benefits usually remain out of reach. Nevertheless, it is theoretically possible to obtain a 0% majority voting error using a finite number of classifiers at error levels lower than 50%. We attempt to show that structuring classifiers into relevant multistage organisations can widen this boundary, as well as the limits of majority voting error, even more. Introducing discrete error distributions for analysis, we show how majority voting errors and their limits depend upon the parameters of a multiple classifier system with hardened binary outputs (correct/incorrect). Moreover, we investigate the sensitivity of boundary distributions of classifier outputs to small discrepancies modelled by the random changes of votes, and propose new more stable patterns of boundary distributions. Finally, we show how organising classifiers into different structures can be used to widen the limits of majority voting errors, and how this phenomenon can be effectively exploited.
|Additional Information:||This paper presents a detailed analysis of the majority voting errors (MVE) and their limits considered in the context of multiple classifier systems (MCS). There are many significant contributions in the paper which include: a) avoiding the exponentially complex calculations of majority voting error by replacing it with analysis of the proposed discrete and approximated-continuous error distributions; b) proposing of stable boundary error distributions preserving the absolute theoretical limits of MVE but at the same time substantially improving the classification margins. The main original finding of the paper, the first of this type for MCS, is possibility of widening the limits of MVE by structuring classifiers into relevant multistage organisations (layers). We provided a formal definition of multistage organisation, derived conditions of its optimality and showed many examples proving our claim that MVE limits for multistage organisations tend to widen and virtually disappear for sufficiently large number of classifiers. This theoretical derivations have attracted interest and led to practical predictive performance gains in our collaborations with major industrial partners from telecommunication and airline industries.|
|Uncontrolled Keywords:||Combining classifiers, error distribution, generalisation, majority voting, margin. multistage organisations|
|Subjects:||Generalities > Computer Science and Informatics > Artificial Intelligence|
Generalities > Computer Science and Informatics
|Group:||School of Design, Engineering & Computing > Smart Technology Research Centre|
|Deposited By:||INVALID USER|
|Deposited On:||17 Dec 2007|
|Last Modified:||07 Mar 2013 14:36|
|Repository Staff Only -|
|BU Staff Only -|
|Help Guide -||Editing Your Items in BURO|