Mixture Regression for Covariate Shift

By Amos J. Storkey, Masashi Sugiyama. 2006

125 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Molecular Biology

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

In Molecular Biology

Citations in the corpus

Transductive Multi-View Zero-Shot Learning (2015, background): “This process alleviates the projection domain shift problem, as well as providing a common space in which heterogeneous views can be directly compared, and their complementarity exploited (Sec.”
Domain Transfer Multiple Kernel Learning (2012, background): “Several cross-domain learning methods [15], [29] were proposed to cope with the inconsistency of data distributions (such as covariate shift [29] or sampling selection bias [15]).”
Adaptive Risk Minimization: A Meta-Learning Approach for Tackling Group Shift (2020, background): “This formulation is a generalization of common settings such as label shift and, arguably, most cases of covariate shift [63, 51, 58].”
Dataset Shift in Machine Learning (2009, methodology): “In the current context, a Bayesian information criterion method can be used [Storkey and Sugiyama, 2007] for selection of the number of mixtures, but it may not always work well as it is both approximate and a heuristic for latent variable problems.”
Learning to Learn Single Domain Generalization (2020, background): “An interesting yet seldom investigated problem then arises: Can a model generalize from one source domain to many unseen target domains? In other words, how to maximize the model generalization when there is only a single domain available for training? The discrepancy between source and target domai”
Unsupervised Domain Adaptation with a Relaxed Covariate Shift Assumption (2017, background): “Covariate shift states that the conditional labeling distributions of both the source and target domains are the same, while their respective marginal data distributions can be different (Storkey and Sugiyama 2006; Ben-David and Urner 2012; 2014).”
Domain Transfer SVM for video concept detection (2009, methodology): “Several crossdomain learning methods [6, 15] were also proposed to cope with inconsistency of data distribution (such as covariate shift [15] or sampling selection bias [6]).”
Limits of Estimating Heterogeneous Treatment Effects: Guidelines for Practical Algorithm Design (2018, methodology): “Previous works tend to adjust for selection bias “mechanically” using variants of importance sampling approaches based on inverse-propensity-weighting (IPW) (Sugiyama et al., 2007; Shimodaira, 2000), and kernel mean matching (Huang et al.”
Augmenting Image Classifiers Using Data Augmentation Generative Adversarial Networks (2018, background): “Transfer Learning and Dataset Shift: The term dataset shift [36] generalises the concept of covariate shift [33, 37, 38] to multiple cases of changes between domains.”
Second opinion needed: communicating uncertainty in medical machine learning (2021, background): “Ultimately, Gaussian processes scale poorly in the number of data points 47 and have been challenging to apply to structured problems where a good covariance function is unknown a priori (i.e. in the case of dataset shift) 24,51 .”
Covariate Shift Adaptation by Importance Weighted Cross Validation (2007, background): “Bayesian generative approaches to coping with such situations have been proposed when unlabeled test input points are available (Storkey and Sugiyama, 2007 ) or when both test input points and test output values are available (Daumé III and Marcu, 2006).”
When Training and Test Sets are Different: Characterising Learning Transfer (2013, methodology): “In the current context, Bayesian Information Criterion methods can be used [23] for selection of the number of mixtures, but it may not always work well as it is both approximate and a heuristic for latent variable problems.”