Modeling Position Specificity in Sequence Kernels by Fuzzy Equivalence Relations
Ulrich Bodenhofer, Karin Schwarzbauer, Mihaela Ionescu, Sepp Hochreiter · 2009
This paper demonstrates that several known sequence kernels can be expressed in a unified framework in which the position specificity is modeled by fuzzy equivalence relations. In addition to this interpretation, we address the practical issues of positive semidefiniteness, computational complexity, and the extraction of interpretable features from the final support vector machine classifier.