Support and invertibility in domain-invariant representations
Name
johansson19a.pdf
Description
Published version
Size
615.35 KB
Format
Adobe PDF
Checksum (MD5)
0358c190bfffe8809dd23456047c3e80
Author(s)
Johansson, Fredrik D.
Sontag, David Alexander
Date Issued
April 2019
Journal
Proceedings of Machine Learning Research
Publisher
International Machine Learning Society
Citation
Johansson, Fredrik D. et al. “Support and invertibility in domain-invariant representations.” Paper in the Proceedings of Machine Learning Research, 89, 22nd International Conference on Artificial Intelligence and Statistics (AISTATS) 2019, Naha,Okinawa, Japa, April 16-18 2019, International Machine Learning Society: 527-536 # © 2019 The Author(s)
Version
Author's final manuscript
Abstract
Learning domain-invariant representations has become a popular approach to unsupervised domain adaptation and is often justified by invoking a particular suite of theoretical results. We argue that there are two significant flaws in such arguments. First, the results in question hold only for a fixed representation and do not account for information lost in non-invertible transformations. Second, domain invariance is often a far too strict requirement and does not always lead to consistent estimation, even under strong and favorable assumptions. In this work, we give generalization bounds for unsupervised domain adaptation that hold for any representation function by acknowledging the cost of non-invertibility. In addition, we show that penalizing distance between densities is often wasteful and propose a bound based on measuring the extent to which the support of the source domain covers the target domain. We perform experiments on well-known benchmarks that illustrate the short-comings of current standard practice.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.
Persistent DSpace Link
DOI of Published Version
http://proceedings.mlr.press/v89/johansson19a.html
