Hierarchical Penalization

Abstract : Hierarchical penalization is a generic framework for incorporating prior information in the fitting of statistical models, when the explicative variables are organized in a hierarchical structure. The penalizer is a convex functional that performs soft selection at the group level, and shrinks variables within each group. This favors solutions with few leading terms in the final combination. The framework, originally derived for taking prior knowledge into account, is shown to be useful in linear regression, when several parameters are used to model the influence of one feature, or in kernel regression, for learning multiple kernels.
Complete list of metadatas

Cited literature [8 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-00267338
Contributor : Marie Szafranski <>
Submitted on : Thursday, March 27, 2008 - 11:35:32 AM
Last modification on : Tuesday, October 23, 2018 - 7:12:03 PM
Long-term archiving on : Friday, September 28, 2012 - 11:50:08 AM

File

NIPS2007_0613.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00267338, version 1

Citation

Marie Szafranski, Yves Grandvalet, Pierre Morizet-Mahoudeaux. Hierarchical Penalization. Neural Information Processing Systems (NIPS), 2007, Canada. pp.1457--1464. ⟨hal-00267338⟩

Share

Metrics

Record views

185

Files downloads

69