Working Set Selection Using Functional Gain for LS-SVM

Liefeng Bo, Licheng Jiao and Ling Wang

Xidian University

Abstract: The efficiency of sequential minimal optimization (SMO) [2] depends strongly on the working set selection. This letter shows how the improvement of SMO in each iteration, named the functional gain (FG) [1], is used to select the working set for least squares support vector machine (LSSVM) [3]. We prove the convergence of the proposed method and give some theoretical support for its performance. Empirical comparisons demonstrate that our method is superior to the maximum violating pair (MVP) working set selection.


  1. Liefeng Bo, Licheng Jiao and Ling Wang, Working Set Selection Using Functional Gain for LS-SVM, IEEE Transactions on Neural Networks (TNN), vol. 18(5), pp. 1541-1544, 2007. [PDF] [BIB]

  2. J. Platt, Sequential Minimal Optimization: A Fast Algorithm for Training Support Vector Machines, Advance in Kernel Methods --- Support Vector Learning, pp. 185-208, 1999.

  3. J. Suykens and J. Vandewalle, Least Squares Support Vector Machine Classifiers, Neural Process Letter, vol. 9, pp. 293-300, 1999.

Matlab and C++ Code

Description: FSALS-SVM is  a package for implementing fast sparse approximation for LS-SVM.

Requirement: Matlab 7.01 and VC++ 6.0.

Download: [code]. This package is free for academic usage. You can run it at your own risk.