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 suboptimal 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. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1109/MLSP.2019.8918731 | 2019 IEEE 29th International Workshop on Machine Learning for Signal Processing (MLSP) |
Keywords | DocType | ISSN |
Sample selection bias,cross-validation | Conference | 1551-2541 |
ISBN | Citations | PageRank |
978-1-7281-0825-4 | 0 | 0.34 |
References | Authors | |
0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Wouter M. Kouw | 1 | 6 | 2.89 |
Jesse H. Krijthe | 2 | 26 | 5.32 |
Marco Loog | 3 | 1796 | 154.31 |