A seminar titled " Multiple Kernel Learning" is held in KAIST.
Title : Multiple Kernel Learning
Date : October 23 (Tuesday), 2012 14:00
Location : Wooribyul Seminar Room(#2201), Department of Electrical Engineering
Speaker : Dr. Maius Kloft/ Berlin Institute of Technology
In real-world applications such as bioinformatics and computer vision, data frequently arises from multiple heterogeneous sources or is represented by various complementary views -- or kernels! --, the right choice of which being unknown. Unfortunately, classical approaches to multiple kernel learning (MKL) are rarely observed to outperform trivial
baselines in practical applications. To this end, we have developed the Lp-norm MKL methodology, which turns out being both more efficient and more accurate than previous approaches to MKL, allowing us to deal with up to ten thousands of data points and thousands of kernels at the same time. Empirical applications of Lp-norm MKL to diverse, challenging problems from the domains of bioinformatics and computer vision show that Lp-norm MKL achieves accuracies that surpass the state-of-the-art. The proposed
techniques are underpinned by deep foundations in the theory of learning: we prove tight lower and upper bounds on the local and global Rademacher complexities of Lp-norm MKL, which yield faster excess risk bounds than previous works.
In Dec 2012, Marius Kloft is starting as a joint postdoc at the Courant Institute of Mathematical Sciences (New York University) and the Memorial Sloan-Kettering Cancer Center, New York. Currently, he is a postdoc at the Machine Learning Laboratory, Berlin Institute of Technology (TU Berlin), headed by Klaus-Robert Müller. From 2007-2011, he was a PhD student in the machine learning program of TU Berlin, being co-advised by Gilles Blanchard and Peter L. Bartlett, whose learning theory group at UC Berkeley I visited from October 2009 to October 2010. In 2006, he received a diploma (MSc equivalent) in mathematics from the University of Marburg with a thesis in algebraic geometry.