On Regularization Parameter Estimation under Covariate Shift

Wouter Kouw; Marco Loog

doi:10.1109/ICPR.2016.7899671

On Regularization Parameter Estimation under Covariate Shift

Pattern Recognition and Bioinformatics

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

1 Citation (Scopus)

Abstract

This paper identifies a problem with the usual procedure for L2-regularization parameter estimation in a domain adaptation setting. In such a setting, there are differences between the distributions generating the training data (source domain) and the test data (target domain). The usual cross-validation
procedure requires validation data, which can not be obtained from the unlabeled target data. The problem is that if one decides to use source validation data, the regularization parameter is underestimated. One possible solution is to scale the source validation data through importance weighting, but we show that
this correction is not sufficient. We conclude the paper with an empirical analysis of the effect of several importance weight estimators on the estimation of the regularization parameter.

Original language	English
Title of host publication	2016 23rd International Conference on Pattern Recognition (ICPR)
Place of Publication	Piscataway, NJ
Publisher	IEEE
Pages	426-431
Number of pages	6
ISBN (Electronic)	978-1-5090-4847-2
ISBN (Print)	978-1-5090-4848-9
DOIs	https://doi.org/10.1109/ICPR.2016.7899671
Publication status	Published - 2016
Event	ICPR 2016: 23rd International Conference on Pattern Recognition - Cancún, Mexico Duration: 4 Dec 2016 → 8 Dec 2016 Conference number: 23

Conference

Conference	ICPR 2016
Country/Territory	Mexico
City	Cancún
Period	4/12/16 → 8/12/16

Keywords

Training
Pattern recognition
Estimation
Risk management
Parameter estimation
Training data
Temperature measurement

Access to Document

10.1109/ICPR.2016.7899671

Cite this

@inproceedings{74825766254c455b82f878247beff40b,

title = "On Regularization Parameter Estimation under Covariate Shift",

abstract = "This paper identifies a problem with the usual procedure for L2-regularization parameter estimation in a domain adaptation setting. In such a setting, there are differences between the distributions generating the training data (source domain) and the test data (target domain). The usual cross-validationprocedure requires validation data, which can not be obtained from the unlabeled target data. The problem is that if one decides to use source validation data, the regularization parameter is underestimated. One possible solution is to scale the source validation data through importance weighting, but we show thatthis correction is not sufficient. We conclude the paper with an empirical analysis of the effect of several importance weight estimators on the estimation of the regularization parameter.",

keywords = "Training, Pattern recognition, Estimation, Risk management, Parameter estimation, Training data, Temperature measurement",

author = "Wouter Kouw and Marco Loog",

year = "2016",

doi = "10.1109/ICPR.2016.7899671",

language = "English",

isbn = "978-1-5090-4848-9 ",

pages = "426--431",

booktitle = "2016 23rd International Conference on Pattern Recognition (ICPR)",

publisher = "IEEE",

address = "United States",

note = "ICPR 2016 : 23rd International Conference on Pattern Recognition ; Conference date: 04-12-2016 Through 08-12-2016",

}

TY - GEN

T1 - On Regularization Parameter Estimation under Covariate Shift

AU - Kouw, Wouter

AU - Loog, Marco

N1 - Conference code: 23

PY - 2016

Y1 - 2016

N2 - This paper identifies a problem with the usual procedure for L2-regularization parameter estimation in a domain adaptation setting. In such a setting, there are differences between the distributions generating the training data (source domain) and the test data (target domain). The usual cross-validationprocedure requires validation data, which can not be obtained from the unlabeled target data. The problem is that if one decides to use source validation data, the regularization parameter is underestimated. One possible solution is to scale the source validation data through importance weighting, but we show thatthis correction is not sufficient. We conclude the paper with an empirical analysis of the effect of several importance weight estimators on the estimation of the regularization parameter.

AB - This paper identifies a problem with the usual procedure for L2-regularization parameter estimation in a domain adaptation setting. In such a setting, there are differences between the distributions generating the training data (source domain) and the test data (target domain). The usual cross-validationprocedure requires validation data, which can not be obtained from the unlabeled target data. The problem is that if one decides to use source validation data, the regularization parameter is underestimated. One possible solution is to scale the source validation data through importance weighting, but we show thatthis correction is not sufficient. We conclude the paper with an empirical analysis of the effect of several importance weight estimators on the estimation of the regularization parameter.

KW - Training

KW - Pattern recognition

KW - Estimation

KW - Risk management

KW - Parameter estimation

KW - Training data

KW - Temperature measurement

U2 - 10.1109/ICPR.2016.7899671

DO - 10.1109/ICPR.2016.7899671

M3 - Conference contribution

SN - 978-1-5090-4848-9

SP - 426

EP - 431

BT - 2016 23rd International Conference on Pattern Recognition (ICPR)

PB - IEEE

CY - Piscataway, NJ

T2 - ICPR 2016

Y2 - 4 December 2016 through 8 December 2016

ER -

On Regularization Parameter Estimation under Covariate Shift

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this