The Planning-ahead SMO Algorithm

TL;DR

This paper introduces a planning-ahead modification to the SMO algorithm for SVM training, improving step size and convergence, with extensive experiments showing its superior performance.

Contribution

It proposes a novel planning-ahead approach to enhance the SMO algorithm's efficiency and convergence guarantees for large-scale SVM training.

Findings

The new algorithm converges reliably to the optimum.

Experiments show improved training speed over standard SMO.

The method performs well across diverse datasets.

Abstract

The sequential minimal optimization (SMO) algorithm and variants thereof are the de facto standard method for solving large quadratic programs for support vector machine (SVM) training. In this paper we propose a simple yet powerful modification. The main emphasis is on an algorithm improving the SMO step size by planning-ahead. The theoretical analysis ensures its convergence to the optimum. Experiments involving a large number of datasets were carried out to demonstrate the superiority of the new algorithm.