===== 12 - Support Vector Machines ======

April 1st, 2010

=== Slides ===

[[cos424>slides/12-svm.pdf|PDF]] [[cos424>slides/12-svm.djvu|DJVU]]   

=== Scribe Notes ===

FIXME

=== Summary ===

   * Maximizing margins.
   * Soft margins.
   * Kernels.
   * Kernels everywhere. 

=== Reading ===

Many researchers have first approached SVMs with Chris Burges' tutorial.
Section 1 to 4 of the Bottou and Lin chapter propose a //self-contained//
mathematical treatment of duality in SVMs.


   * (optional) C. J. C. Burges: [[http://research.microsoft.com/en-us/um/people/cburges/papers/svmtutorial.pdf|A Tutorial on Support Vector Machines for Pattern Recognition]]. 1998.

   * (optional) L. Bottou and C.-J. Lin: [[http://leon.bottou.org/papers/bottou-lin-2006|Support Vector Machine Solvers]]. In //Large Scale Kernel Machines//, MIT Press, 2007.  (section 1-4)

=== Software ===

The most used softwares for SVMs 
are [[http://www.csie.ntu.edu.tw/~cjlin/libsvm|LibSVM]] 
and [[http://svmlight.joachims.org/|SVMLight]].
Also the [[http://leon.bottou.org/projects/svqp|SVQP2]] solver is
a small and efficient C++ library for solving the SVM dual optimization problem.
Note that one should use very different algorithms for training SVMs with kernels
and SVMs with linear parametrization.