I’m a first year PhD student of Tula State University. My doctoral research investigates to the high-performance solution for large-scale recovering dependences. My work proposes a simple approach to finding a solution of the SVM problem, based on averaging the decision rules, which are constructed from small random, possibly intersecting subsamples of the initial training set. The proposed approach allows to quickly find an approximate (but not very different from the exact) solution of the SVM problem, even for large training sets.