Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

Robust importance-weighted cross-validation under sample selection bias

Abstract: 

Cross-validation under sample selection bias can, in principle, be done by importance-weighting the empirical risk. However, the importance-weighted risk estimator produces sub-optimal hyperparameter estimates in problem settings where large weights arise with high probability. We study its sampling variance as a function of the training data distribution and introduce a control variate to increase its robustness to problematically large weights.

up
0 users have voted:

Paper Details

Authors:
Wouter M Kouw, Jesse H Krijthe, Marco Loog
Submitted On:
11 October 2019 - 4:48pm
Short Link:
Type:
Poster
Event:
Presenter's Name:
Wouter Kouw
Paper Code:
27
Document Year:
2019
Cite

Document Files

PDF of poster

(17)

Subscribe

[1] Wouter M Kouw, Jesse H Krijthe, Marco Loog, "Robust importance-weighted cross-validation under sample selection bias", IEEE SigPort, 2019. [Online]. Available: http://sigport.org/4857. Accessed: Nov. 21, 2019.
@article{4857-19,
url = {http://sigport.org/4857},
author = {Wouter M Kouw; Jesse H Krijthe; Marco Loog },
publisher = {IEEE SigPort},
title = {Robust importance-weighted cross-validation under sample selection bias},
year = {2019} }
TY - EJOUR
T1 - Robust importance-weighted cross-validation under sample selection bias
AU - Wouter M Kouw; Jesse H Krijthe; Marco Loog
PY - 2019
PB - IEEE SigPort
UR - http://sigport.org/4857
ER -
Wouter M Kouw, Jesse H Krijthe, Marco Loog. (2019). Robust importance-weighted cross-validation under sample selection bias. IEEE SigPort. http://sigport.org/4857
Wouter M Kouw, Jesse H Krijthe, Marco Loog, 2019. Robust importance-weighted cross-validation under sample selection bias. Available at: http://sigport.org/4857.
Wouter M Kouw, Jesse H Krijthe, Marco Loog. (2019). "Robust importance-weighted cross-validation under sample selection bias." Web.
1. Wouter M Kouw, Jesse H Krijthe, Marco Loog. Robust importance-weighted cross-validation under sample selection bias [Internet]. IEEE SigPort; 2019. Available from : http://sigport.org/4857