Globally sparse PLS regression

Abstract : Partial least squares (PLS) regression combines dimensionality reduction and prediction using a latent variable model. It provides better predictive ability than principle component analysis by taking into account both the independent and re- sponse variables in the dimension reduction procedure. However, PLS suffers from over-fitting problems for few samples but many variables. We formulate a new criterion for sparse PLS by adding a structured sparsity constraint to the global SIMPLS optimization. The constraint is a sparsity-inducing norm, which is useful for selecting the important variables shared among all the components. The optimization is solved by an augmented Lagrangian method to obtain the PLS components and to perform variable selection simultaneously. We propose a novel greedy algorithm to overcome the computation difficulties. Experiments demonstrate that our approach to PLS regression attains better performance with fewer selected predictors
Complete list of metadatas

Cited literature [18 references]  Display  Hide  Download

https://hal-supelec.archives-ouvertes.fr/hal-01069009
Contributor : Alexandra Siebert <>
Submitted on : Friday, September 26, 2014 - 4:09:48 PM
Last modification on : Thursday, October 10, 2019 - 7:24:09 AM
Long-term archiving on : Friday, April 14, 2017 - 2:44:51 PM

File

liu_GSIMPLS_Springer13.pdf
Files produced by the author(s)

Identifiers

Collections

Citation

Tzu-Yu Liu, Laura Trinchera, Arthur Tenenhaus, Dennis Wei, Alfred Hero. Globally sparse PLS regression. New perspectives in Partial Least Squares and Related Methods, Springer, pp.117-127, 2013, Springer Proceedings in Mathematics & Statistics, ⟨10.1007/978-1-4614-8283-3_7⟩. ⟨hal-01069009⟩

Share

Metrics

Record views

284

Files downloads

484