Covariate Shift Adaptation by Importance Weighted Cross Validation

By Masashi Sugiyama, Matthias Krauledat, K. Müller. 2007

1,089 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Security analysis of online centroid anomaly detection (2010, background): “The most important application field in which robustness of learning algorithms against adversarial input is crucial is computer security.”
Learning explanations that are hard to vary (2020, background): “Covariate Shift [Sugiyama et al., 2007, Quionero-Candela et al., 2009, Sugiyama and Kawanabe, 2012] formalizes the case in which the distribution of the covariates at test time differs from that observed at training.”
Counterfactual reasoning and learning systems: the example of computational advertising (2012, background): “Importance sampling is widely used to deal with covariate shifts (Shimodaira, 2000; Sugiyama et al., 2007).”
How to Explain Individual Classification Decisions (2009, methodology): “For further approaches to data set shift see Sugiyama, Nakajima, Kashima, von Buenau, and Kawanabe (2007b), Sugiyama, Krauledat, and M¨uller (2007a) and the book by Quionero-Candela, Sugiyama, Schwaighofer, and Lawrence (2009).”
Introduction to machine learning for brain imaging (2011, methodology): “Moreover, a new generation of source localization techniques is now in use (e.g. Haufe et al., 2008, 2009) and has been successfully applied in BCI research (e.g. Pun et al., 2006; Noirhomme et al., 2008; Grosse-Wentrup et al., 2007).”
Using Stacking to Average Bayesian Predictive Distributions (with Discussion) (2017, background): “If we assume that the future distribution is different from the past, then the covariate shift can be taken into account by weighting (e.g. Shimodaira, 2000; Sugiyama and Müller, 2005; Sugiyama et al., 2007).”
Localized Conformal Prediction: A Generalized Inference Framework for Conformal Prediction (2021, background): “(2019a), the authors consider average coverage for the test sample distribution, given that we have enough test samples to estimate the covariate shift (Shimodaira, 2000; Sugiyama and Müller, 2005; Sugiyama et al., 2007; Quionero-Candela et al., 2009).”
Approximate inference for the loss-calibrated Bayesian (2011, background): “We could thus handle the covariate shift problem (Sugiyama et al., 2007) with a set of unlabelled examples from the test distribution.”
Learning Non-stationary System Dynamics Online Using Gaussian Processes (2010, methodology): “This procedure is also known as importance-weighted cross-validation [11].”
Identifying Best Interventions through Online Importance Sampling (2017, methodology): “In the context of machine learning, importance sampling has been mostly used to recondition input data to adhere to conditions imposed by learning algorithms (Sugiyama et al., 2007; Li et al., 2011; Zhang & Zhao, 2014).”
Semi-Supervised Learning of Class Balance under Class-Prior Change by Distribution Matching (2012, methodology): “The width of the Gaussian kernel K(x,x′) and the regularization parameter δ(≥ 0) are chosen by 5-fold weighted cross-validation (Sugiyama et al., 2007) in terms of the misclassification error.”
A Brief Review of Domain Adaptation (2020, methodology): “Kullback-Leibler Importance Estimation Procedure (KLIEP) [41], [42], [43], directly estimates the density ratio using Kl-divergence between target distribution and weighted source distribution.”
Transfer Learning in Brain-Computer Interfaces (2015, methodology): “have used covariate shift adaptation to combine labeled training data with unlabeled test data [26].”
Evaluating Prediction-Time Batch Normalization for Robustness under Covariate Shift (2020, background): “Importance weighting methods are also a common approach, where training examples are reweighted to minimize an estimate of the predictive risk in the target domain [Sugiyama et al., 2007].”
Learning Equations for Extrapolation and Control (2018, critical): “However, existing domain adaptation techniques, such as sample reweighting (Sugiyama et al., 2007) are not applicable, because training and extrapolation domain are disjoint for us.”