Cargando…

Robust priors for regularized regression

Induction benefits from useful priors. Penalized regression approaches, like ridge regression, shrink weights toward zero but zero association is usually not a sensible prior. Inspired by simple and robust decision heuristics humans use, we constructed non-zero priors for penalized regression models...

Descripción completa

Detalles Bibliográficos
Autores principales: Bobadilla-Suarez, Sebastian, Jones, Matt, Love, Bradley C.
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Elsevier 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8903146/
https://www.ncbi.nlm.nih.gov/pubmed/34861584
http://dx.doi.org/10.1016/j.cogpsych.2021.101444
_version_ 1784664706482438144
author Bobadilla-Suarez, Sebastian
Jones, Matt
Love, Bradley C.
author_facet Bobadilla-Suarez, Sebastian
Jones, Matt
Love, Bradley C.
author_sort Bobadilla-Suarez, Sebastian
collection PubMed
description Induction benefits from useful priors. Penalized regression approaches, like ridge regression, shrink weights toward zero but zero association is usually not a sensible prior. Inspired by simple and robust decision heuristics humans use, we constructed non-zero priors for penalized regression models that provide robust and interpretable solutions across several tasks. Our approach enables estimates from a constrained model to serve as a prior for a more general model, yielding a principled way to interpolate between models of differing complexity. We successfully applied this approach to a number of decision and classification problems, as well as analyzing simulated brain imaging data. Models with robust priors had excellent worst-case performance. Solutions followed from the form of the heuristic that was used to derive the prior. These new algorithms can serve applications in data analysis and machine learning, as well as help in understanding how people transition from novice to expert performance.
format Online
Article
Text
id pubmed-8903146
institution National Center for Biotechnology Information
language English
publishDate 2022
publisher Elsevier
record_format MEDLINE/PubMed
spelling pubmed-89031462022-03-11 Robust priors for regularized regression Bobadilla-Suarez, Sebastian Jones, Matt Love, Bradley C. Cogn Psychol Article Induction benefits from useful priors. Penalized regression approaches, like ridge regression, shrink weights toward zero but zero association is usually not a sensible prior. Inspired by simple and robust decision heuristics humans use, we constructed non-zero priors for penalized regression models that provide robust and interpretable solutions across several tasks. Our approach enables estimates from a constrained model to serve as a prior for a more general model, yielding a principled way to interpolate between models of differing complexity. We successfully applied this approach to a number of decision and classification problems, as well as analyzing simulated brain imaging data. Models with robust priors had excellent worst-case performance. Solutions followed from the form of the heuristic that was used to derive the prior. These new algorithms can serve applications in data analysis and machine learning, as well as help in understanding how people transition from novice to expert performance. Elsevier 2022-02 /pmc/articles/PMC8903146/ /pubmed/34861584 http://dx.doi.org/10.1016/j.cogpsych.2021.101444 Text en © 2021 The Author(s) https://creativecommons.org/licenses/by/4.0/This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
spellingShingle Article
Bobadilla-Suarez, Sebastian
Jones, Matt
Love, Bradley C.
Robust priors for regularized regression
title Robust priors for regularized regression
title_full Robust priors for regularized regression
title_fullStr Robust priors for regularized regression
title_full_unstemmed Robust priors for regularized regression
title_short Robust priors for regularized regression
title_sort robust priors for regularized regression
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8903146/
https://www.ncbi.nlm.nih.gov/pubmed/34861584
http://dx.doi.org/10.1016/j.cogpsych.2021.101444
work_keys_str_mv AT bobadillasuarezsebastian robustpriorsforregularizedregression
AT jonesmatt robustpriorsforregularizedregression
AT lovebradleyc robustpriorsforregularizedregression