Linear convergence rate for the MDM algorithm for the Nearest Point Problem

作者：

Highlights：

• Linear convergence rate for the MDM algorithm is proved.

• No particular assumption is needed for the proof.

• A geometric and simple approach is followed.

• This may pave the way for proving similar results for the SMO algorithm used to train SVMs.

摘要

Highlights•Linear convergence rate for the MDM algorithm is proved.•No particular assumption is needed for the proof.•A geometric and simple approach is followed.•This may pave the way for proving similar results for the SMO algorithm used to train SVMs.

论文关键词：Convergence,MDM algorithm,Nearest Point Problem,Convex Hulls,Support vector machines,ν-SVMs

论文评审过程：Received 28 February 2014, Revised 4 September 2014, Accepted 11 October 2014, Available online 22 October 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.10.015