Maximum-Gain Working Set Selection for SVMs

  title={Maximum-Gain Working Set Selection for SVMs},
  author={Tobias Glasmachers and Christian Igel},
  journal={Journal of Machine Learning Research},
Support vector machines are trained by solving constrained qua ratic optimization problems. This is usually done with an iterative decomposition algorithm o perating on a small working set of variables in every iteration. The training time strongly depend s on the selection of these variables. We propose the maximum-gain working set selection algorithm f or large scale quadratic programming. It is based on the idea to greedily maximize the progress in ea ch single iteration. The algorithm takes… CONTINUE READING
43 Citations
9 References
Similar Papers

Similar Papers

Loading similar papers…