The HPSVM Procedure

The support vector machines (SVM) algorithm is popular in the data mining area of classification. The HPSVM procedure executes the SVM algorithm in high-performance mode, which enables you to run the procedure on multiple threads in single-machine mode or, if you have a license for SAS High-Performance Data Mining, in multiple threads on multiple machines (distributed mode). PROC HPSVM uses both linear and nonlinear kernels to conduct training. Like other high-performance procedures, the HPSVM procedure can read data in distributed form, and it can perform computation in parallel in either single-machine or distributed mode. PROC HPSVM provides two optimization techniques: the interior-point method and the active-set method. The interior-point method can run in either single-machine mode or distributed mode, whereas active-set method runs only in single-machine mode.