===== Trading Convexity for Scalability =====

//Abstract//:
Convex learning algorithms, such as Support Vector Machines (SVMs), are
often seen as highly desirable because they offer strong practical
properties and are amenable to theoretical analysis.  However, in this work
we show how non-convexity can provide scalability advantages over
convexity.  We show how concave-convex programming can be applied to produce
(i) faster SVMs where training errors are no longer support vectors, and
(ii) much faster Transductive SVMs.


<box 99% orange>
Ronan Collobert, Jason Weston and Léon Bottou: **Trading Convexity for Scalability**,  //Proceedings of the Twenty-third International Conference on Machine Learning (ICML 2006)//, IMLS/ICML, 2006.

[[http://leon.bottou.org/publications/djvu/icml-2006a.djvu|icml-2006a.djvu]]
[[http://leon.bottou.org/publications/pdf/icml-2006a.pdf|icml-2006a.pdf]]
[[http://leon.bottou.org/publications/psgz/icml-2006a.ps.gz|icml-2006a.ps.gz]]
</box>

  @inproceedings{collobert-weston-bottou-2006,
    author = {Collobert, Ronan and Weston, Jason and Bottou, L\'{e}on},
    title = {Trading Convexity for Scalability},
    year = {2006},
    booktitle = {Proceedings of the Twenty-third International Conference on Machine Learning (ICML 2006)},
    publisher = {IMLS/ICML},
    note = {ACM Digital Library},
    url = {http://leon.bottou.org/papers/collobert-weston-bottou-2006},
  }


==== Links ====

  * The [[http://www.kyb.tuebingen.mpg.de/bs/people/fabee/universvm.html|UniverSVM]] software is based on a modified version of the [[:projects:svqp|SVQP2]] optimizer.
  * About the DC algorithms [[http://lita.sciences.univ-metz.fr/~lethi/DCA.html]]
  * Home page of [[http://www.kyb.tuebingen.mpg.de/~weston|Jason Weston]]
  * Home page of [[http://ronan.collobert.com/|Ronan Collobert]]
  * Home page of [[http://www.kyb.tuebingen.mpg.de/~fabee|Fabian Sinz]]