A Cascade Method for Reducing Training Time and the Number of Support Vectors

摘要

A novel cascade learning strategy for training support vector machines (SVMs) is proposed to speed up the training of SVMs. The training procedure consists of three steps which are performed in a cascade way. All the subproblems are processed parallelly in each step, and non-support-vector data are filtered out step by step. The simulation results indicate that our method not only speeds up the training procedure while maintaining the generalization accuracy of SVMs but also reduces the number of support vectors.