Combining Support Vector Machines by Means of Fuzzy Aggregation
Martin Hole, Jaroslav Moravec · 2007
The paper deals with a recently proposed approach to combining classifiers by means of fuzzy ag- gregation. The approach relies on the quasi-Sugeno integral and on the t-conorm integral as a generalization of the Choquet and Sugeno integral, which have been used for combining classifiers so far. New theoretical devel- opment is presented, in particular a proposition concerning the measures used in the quasi-Sugeno integral, and the approach is elaborated specifically for support vector machines. Finally, experience is reported that was gained when using the approach to combine support vector machines in a neurophysiologic application.