Multiple Kernel Learning Methods and their Application in Yeast Protein Subcellular Localization Prediction

Abstract

Kernel methods are becoming more and more popular technique for solving machine learning problems. Recent advances in the field of Multiple Kernel Learning (MKL) have highlighted MKL as an attractive tool that can be applied in many supervised learning tasks. During the past decade, it has been shown that classifiers that use combinations of multiple kernels instead of classical single kernel-based ones attain significantly better results in certain problems.