TR Announcement

[Chiranjib Bhattacharya]

Technical Report Announcement:

Platt's SMO algorithm is an excellent algorithm for designing
SVMs because it is very efficient. It has become popular since
it is also extremely simple to implement. Our recent research
has shown that there is an important source of inefficiency in
the SMO algorithm  that has to do with choosing the threshold 
parameter, b. We remove this inefficiency by carefully modifying
SMO while keeping the main ideas of SMO in tact. The modified
algorithms run much faster than the original SMO. Details are 
given in the Technical Report mentioned below. A gzipped post-
script file containing the report can be downloaded from:
        http://guppy.mpe.nus.edu.sg/~mpessk/
Send any comments to: mpessk at guppy.mpe.nus.edu.sg

----------------------------------------------------------------------------
 Improvements to Platt's SMO Algorithm for SVM Classifier Design

                 Technical Report CD-99-14

  S.S. Keerthi, S.K. Shevade, C. Bhattacharyya &  K.R.K. Murthy 


                         Abstract
      
This  paper  points out an important source of confusion and inefficiency
in Platt's  Sequential Minimal Optimization (SMO) algorithm that  is caused  
by the use of a single threshold value. Using clues from the KKT conditions 
for  the dual problem, two  threshold  parameters are  employed to derive 
modifications of SMO. These modified algorithms perform significantly faster 
than the original SMO on all benchmark datasets tried.

----------------------------------------------------------------------------