Semi-generative modelling

Covariate-shift adaptation with cause and effect features

Journal Article (2020)
Author(s)

Julius von Kügelgen (Max Planck Institute for Intelligent Systems, University of Cambridge)

Alexander Mey (Student TU Delft)

Marco Loog (TU Delft - Pattern Recognition and Bioinformatics, University of Copenhagen)

Research Group
Pattern Recognition and Bioinformatics
Copyright
© 2020 Julius von Kügelgen, Alexander Mey, M. Loog
More Info
expand_more
Publication Year
2020
Language
English
Copyright
© 2020 Julius von Kügelgen, Alexander Mey, M. Loog
Research Group
Pattern Recognition and Bioinformatics
Volume number
89
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

Current methods for covariate-shift adaptation use unlabelled data to compute importance weights or domain-invariant features, while the final model is trained on labelled data only. Here, we consider a particular case of covariate shift which allows us also to learn from unlabelled data, that is, combining adaptation with semi-supervised learning. Using ideas from causality, we argue that this requires learning with both causes, XC, and effects, XE, of a target variable, Y, and show how this setting leads to what we call a semi-generative model, P(Y,XE|XC,θ). Our approach is robust to domain shifts in the distribution of causal features and leverages unlabelled data by learning a direct map from causes to effects. Experiments on synthetic data demonstrate significant improvements in classification over purely-supervised and importance-weighting baselines.

Files

Kugelgen19a.pdf
(pdf | 1.05 Mb)
License info not available