Research Article

Cascade Support Vector Machines with Dimensionality Reduction

Algorithm 1

Pseudocode of ECM algorithm. In each cascade level, the -dimensional patterns , which are the pendants of the -dimensional support vectors of the previous cascade level, are employed as training patterns.
Require: training set , ,
()   cv/grid search on
()  repeat
()   for    to    do
()    select training set
()    reduce dimensionality
()    get support vectors
()    train SVM on
()     (pendants of )
()   end for
() 
() until  
() train SVM on
() return  , SVM