Sparse Kernel Orthonormalized PLS for feature extraction in large data sets

Part of Advances in Neural Information Processing Systems 19 (NIPS 2006)

Bibtex Metadata Paper


Jerónimo Arenas-garcía, Kaare Petersen, Lars K. Hansen


In this paper we are presenting a novel multivariate analysis method. Our scheme is based on a novel kernel orthonormalized partial least squares (PLS) variant for feature extraction, imposing sparsity constrains in the solution to improve scalability. The algorithm is tested on a benchmark of UCI data sets, and on the analysis of integrated short-time music features for genre prediction. The upshot is that the method has strong expressive power even with rather few features, is clearly outperforming the ordinary kernel PLS, and therefore is an appealing method for feature extraction of labelled data.