Maximum-Gain Working Set Selection for SVMs

Support vector machines are trained by solving constrained quadratic optimization problems. This is usually done with an iterative decomposition algorithm operating on a small working set of variab...