Direct kernel least-squares support vector machines with heuristic regularization

This tutorial paper introduces direct kernel least squares support vector machines, where traditional ridge regression is applied directly on the kernel transformed data, rather than using the primal dual formulation. A direct kernel method can be any regression model, where the kernel is considered as a data pre-processing step. The emphasis of the paper is that such direct kernel methods often require kernel centering in order to work. A heuristic formula for the regularization parameter is proposed based on preliminary scaling experiments.