I used the normalization operator, using the z-transformation on a data set consisting of 1700 examples and 5000 features (sparse formatted ). The attributes are all integers, stored in the sparse-float-array. The normalization works fine but took a very long time ~ around 20 minutes; the next stage applied learning of an svm which only took a few minutes.
Is there any way to speed up the normalization? I want to apply it to larger data sets in the near future: ~50,000 examples