Binary classification using Support Vector Machines (SVMs) has been an important recent advance in the field of machine learning. This talk gives an introduction to SVMs and the highly structured convex quadratic program that forms the training phase. As the Hessian matrix is completely dense, state-of-the-art approaches have used iterative decomposition algorithms, with Interior Point Methods being used only for the smaller sub-problems. This talk will describe a Separable QP formulation which allows IPMs to solve the SVM training phase with large-scale data sets.
Current 2019 2018 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996
