Multiplicative updates For Non-Negative Kernel SVM

TL;DR

This paper introduces multiplicative update algorithms for non-negative kernel SVMs that converge rapidly and achieve comparable generalization performance to standard SVMs without requiring parameter tuning.

Contribution

It extends multiplicative updates to non-negative kernel SVMs, providing a parameter-free method with proven convergence rates and competitive accuracy.

Findings

Rapid convergence demonstrated in experiments

Achieves similar generalization errors to standard SVMs

Effective across various datasets and non-negative kernels

Abstract

We present multiplicative updates for solving hard and soft margin support vector machines (SVM) with non-negative kernels. They follow as a natural extension of the updates for non-negative matrix factorization. No additional param- eter setting, such as choosing learning, rate is required. Ex- periments demonstrate rapid convergence to good classifiers. We analyze the rates of asymptotic convergence of the up- dates and establish tight bounds. We test the performance on several datasets using various non-negative kernels and report equivalent generalization errors to that of a standard SVM.