<XML><RECORDS><RECORD><REFERENCE_TYPE>3</REFERENCE_TYPE><REFNUM>9161</REFNUM><AUTHORS><AUTHOR>Ying,Y.</AUTHOR><AUTHOR>Campbell,C.</AUTHOR><AUTHOR>Damoulas,T.</AUTHOR><AUTHOR>Girolami,M.A.</AUTHOR></AUTHORS><YEAR>2009</YEAR><TITLE>Class Prediction from Disparate Biological Data Sources using an Iterative Multi-kernel Algorithm</TITLE><PLACE_PUBLISHED>Lecture Notes in Bioinformatics, Proceedings of the 4th IAPR International Conference, Pattern Recognition in Bioinformatics 2009 (PRIB 2009)</PLACE_PUBLISHED><PUBLISHER>Springer Verlag</PUBLISHER><PAGES>427-438</PAGES><LABEL>Ying:2009:9161</LABEL><KEYWORDS><KEYWORD>Multiple kernel learning</KEYWORD></KEYWORDS<ABSTRACT>For many biomedical modelling tasks a number of di?erent types of data may in?uence predictions made by the model. An estab- lished approach to pursuing supervised learning with multiple types of data is to encode these di?erent types of data into separate kernels and use multiple kernel learning. In this paper we propose a simple iterative approach to multiple kernel learning (MKL), focusing on multi-class clas- si?cation. This approach uses a block L1 -regularization term leading to a jointly convex formulation. It solves a standard multi-class classi?cation problem for a single kernel, and then updates the kernel combinatorial coe?cients based on mixed RKHS norms. As opposed to other MKL ap- proaches, our iterative approach delivers a largely ignored message that MKL does not require sophisticated optimization methods while keeping competitive training times and accuracy across a variety of problems. We show that the proposed method outperforms state-of-the-art results on an important protein fold prediction dataset and gives competitive performance on a protein subcellular localization task.</ABSTRACT></RECORD></RECORDS></XML>