2026-09-19T00:23:17Zhttps://dspace.mit.edu/server/oai/requestoai:dspace.mit.edu:1721.1/1391502022-01-15T04:05:02Zcom_1721.1_7582com_1721.1_7581col_1721.1_131023
Jegelka, Stefanie
Torralba, Antonio
Chuang, Ching-Yao
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
2022-01-14T14:52:54Z
2022-01-14T14:52:54Z
2021-06
2021-06-24T19:20:15.437Z
https://hdl.handle.net/1721.1/139150
When the test distribution differs from the training distribution, machine learning models can perform poorly and wrongly overestimate their performance. In this work, we aim to better estimate the model’s performance under distribution shift, without supervision. To do so, we use a set of domain-invariant predictors as a proxy for the unknown, true target labels, where the error of this estimation is bounded by the target risk of the proxy model. Therefore, we study the generalization of domain-invariant representations and show that the complexity of the latent representation has a significant influence on the target risk. Empirically, our estimation approach can self-tune to find the optimal model complexity and the resulting models achieve good target generalization, and estimate target error of other models well. Applications of our results include model selection, deciding early stopping, error detection, and predicting the adaptability of a model between domains.
S.M.
Massachusetts Institute of Technology
In Copyright - Educational Use Permitted
Copyright MIT
http://rightsstatements.org/page/InC-EDU/1.0/
Understanding and Estimating the Adaptability of Domain-Invariant Representations
Thesis
application/pdf
Master
Master of Science in Electrical Engineering and Computer Science
Publication
unknown
unknown
<Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="180709c4-c3a9-4f40-b5ee-6fb24fa54099">
<Type xmlns="https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types">http://purl.org/coar/resource_type/c_1843</Type>
<Title>Understanding and Estimating the Adaptability of Domain-Invariant Representations</Title>
<PublishedIn>
<Publication>
</Publication>
</PublishedIn>
<PublicationDate>2021-06</PublicationDate>
<Authors>
<Author>
<DisplayName>Chuang, Ching-Yao</DisplayName>
<Affiliation>
<OrgUnit>
</OrgUnit>
</Affiliation>
</Author>
</Authors>
<Editors>
</Editors>
<Publishers>
<Publisher>
<DisplayName>Massachusetts Institute of Technology</DisplayName>
<OrgUnit />
</Publisher>
</Publishers>
<License>http://rightsstatements.org/page/InC-EDU/1.0/</License>
<Abstract>When the test distribution differs from the training distribution, machine learning models can perform poorly and wrongly overestimate their performance. In this work, we aim to better estimate the model’s performance under distribution shift, without supervision. To do so, we use a set of domain-invariant predictors as a proxy for the unknown, true target labels, where the error of this estimation is bounded by the target risk of the proxy model. Therefore, we study the generalization of domain-invariant representations and show that the complexity of the latent representation has a significant influence on the target risk. Empirically, our estimation approach can self-tune to find the optimal model complexity and the resulting models achieve good target generalization, and estimate target error of other models well. Applications of our results include model selection, deciding early stopping, error detection, and predicting the adaptability of a model between domains.</Abstract>
<Access xmlns="http://purl.org/coar/access_right"
>
</Access>
</Publication>