An Efficient Method for Large Scale Classification Problems

The support vector machines have been promising methods for classification because of their solid mathematical foundation. However they are not favored for large-scale because the training complexity of SVM is highly dependent on the size of data set.This paper uses Incremental Reduced Support Vector Machines (IRSVM) which begins with an extremely small reduced set and incrementally expands the reduced set based on a series of solved small least squares problems. The size of reduced set is determined dynamically according the algorithm. The experiments on the classification of considerations effecting on printing press progress which is a typical large scale classification problem shows IRSVM has a good efficiency for adjustable printing factor, and computational times as well as memory usage are much smaller for IRSVM than that of conventional SVM.