Domain adaptation algorithms are designed to minimize the misclassification risk of a discriminative model for a target domain with little training data by adapting a model from a source domain with a large amount of training data. Standard approaches measure the adaptation discrepancy based on distance measures between the empirical probability distributions in the source and target domain. In this setting, we address the problem of deriving generalization bounds under practice-oriented general conditions on the underlying probability distributions. As a result, we obtain generalization bounds for domain adaptation based on finitely many moments and smoothness conditions.
%0 Journal Article
%1 Zellinger2021
%A Zellinger, Werner
%A Moser, Bernhard A.
%A Saminger-Platz, Susanne
%D 2021
%J Annals of Mathematics and Artificial Intelligence
%K bernhard cmd deep-learning mappings neural-networks werner
%N 3
%P 333--369
%R 10.1007/s10472-020-09719-x
%T On generalization in moment-based domain adaptation
%U https://link.springer.com/article/10.1007/s10472-020-09719-x
%V 89
%X Domain adaptation algorithms are designed to minimize the misclassification risk of a discriminative model for a target domain with little training data by adapting a model from a source domain with a large amount of training data. Standard approaches measure the adaptation discrepancy based on distance measures between the empirical probability distributions in the source and target domain. In this setting, we address the problem of deriving generalization bounds under practice-oriented general conditions on the underlying probability distributions. As a result, we obtain generalization bounds for domain adaptation based on finitely many moments and smoothness conditions.
@article{Zellinger2021,
abstract = {Domain adaptation algorithms are designed to minimize the misclassification risk of a discriminative model for a target domain with little training data by adapting a model from a source domain with a large amount of training data. Standard approaches measure the adaptation discrepancy based on distance measures between the empirical probability distributions in the source and target domain. In this setting, we address the problem of deriving generalization bounds under practice-oriented general conditions on the underlying probability distributions. As a result, we obtain generalization bounds for domain adaptation based on finitely many moments and smoothness conditions.},
added-at = {2023-08-04T09:45:44.000+0200},
author = {Zellinger, Werner and Moser, Bernhard A. and Saminger-Platz, Susanne},
biburl = {https://www.bibsonomy.org/bibtex/2e05c7b05da076821fd4997d9dfcefdb0/scch},
day = 01,
doi = {10.1007/s10472-020-09719-x},
interhash = {8b518d0a1c6eb39886e6460ff4f5c74e},
intrahash = {e05c7b05da076821fd4997d9dfcefdb0},
issn = {1573-7470},
journal = {Annals of Mathematics and Artificial Intelligence},
keywords = {bernhard cmd deep-learning mappings neural-networks werner},
month = mar,
number = 3,
pages = {333--369},
timestamp = {2023-08-04T09:45:44.000+0200},
title = {On generalization in moment-based domain adaptation},
url = {https://link.springer.com/article/10.1007/s10472-020-09719-x},
volume = 89,
year = 2021
}