Kernel Springy Discriminant Analysis and its Application
to a Phonological Awareness Teaching System

András Kocsor, Kornél Kovács

Research Group on Artificial Intelligence, University of Szeged

Making use of the ubiquitous kernel notion, we present a new nonlinear supervised feature extraction technique called Kernel Springy Discriminant Analysis (KSDA). We demonstrate that this method can efficiently reduce the number of features and increase classification performance.