A Novel SVM Algorithm and Experiment

Standard SVM training has O(m³) time and O(m²) space complexities, where m is the training set size. It is thus computationally infeasible on very large data sets. The author first reviewed the standard minimum enclosing ball (MEB) problems in computational geometry and presented the extensions of minimum enclosing ball problem, Then proposed a novel SVM algorithm-extension core vector machine algorithm (ECVM), which can be used with nonlinear kernels and has a time complexity that is linear in m and a space complexity that is independent of m. Experiment on large data sets-MIT Face Data and Extension demonstrate that the ECVM is as accurate as existing SVM implementations, but is much faster and can handle much larger data sets than existing scale-up methods.