Quadratic programming and penalized regression

Smith, Andrew D. A. C.

Abstract

Quadratic programming is a versatile tool for calculating estimates in penalized regression. It can be used to produce estimates based on L1 roughness penalties, as in total variation denoising. In particular, it can calculate estimates when the roughness penalty is the total variation of a derivative of the estimate. Combining two roughness penalties, the total variation and total variation of the third derivative, results in an estimate with continuous second derivative but controls the number of spurious local extreme values. A multiresolution criterion may be included in a quadratic program to achieve local smoothing without having to specify smoothing parameters. Copyright © Taylor & Francis Group, LLC.

Citation

Smith, A. D. A. C. (2013). Quadratic programming and penalized regression. Communications in Statistics - Theory and Methods, 42(7), 1363-1372. https://doi.org/10.1080/03610926.2012.732177

• Journal Article Type: Conference Paper
• Acceptance Date: Sep 17, 2012
• Online Publication Date: Feb 26, 2013
• Publication Date: May 27, 2013
• Deposit Date: Dec 2, 2015
• Publicly Available Date: Aug 18, 2016
• Journal: Communications in Statistics - Theory and Methods
• Print ISSN: 0361-0926
• Electronic ISSN: 1532-415X
• Publisher: Taylor & Francis
• Peer Reviewed: Peer Reviewed
• Volume: 42
• Issue: 7
• Pages: 1363-1372
• DOI: https://doi.org/10.1080/03610926.2012.732177
• Keywords: multiresolution, nonparametric regression, penalized regression, quadratic programming, total variation
• Public URL: https://uwe-repository.worktribe.com/output/933017
• Publisher URL: http://dx.doi.org/10.1080/03610926.2012.732177
• Additional Information: Additional Information : This is an Accepted Manuscript of an article published by Taylor & Francis in Communications in Statistics - Theory and Methods on 26 February 2013, available online: http://www.tandfonline.com/10.1080/03610926.2012.732177