Mixture Data for Training Cannot Ensure Out-of-distribution Generalization
Mixture Data for Training Cannot Ensure Out-of-distribution Generalization
About this item
Full title
Author / Creator
Zhang, Songming , Luo, Yuxiao , Wang, Qizhou , Haoang Chi , Chen, Xiaofeng , Han, Bo and Li, Jinyan
Publisher
Ithaca: Cornell University Library, arXiv.org
Journal title
Language
English
Formats
Publication information
Publisher
Ithaca: Cornell University Library, arXiv.org
Subjects
More information
Scope and Contents
Contents
Deep neural networks often face generalization problems to handle out-of-distribution (OOD) data, and there remains a notable theoretical gap between the contributing factors and their respective impacts. Literature evidence from in-distribution data has suggested that generalization error can shrink if the size of mixture data for training increas...
Alternative Titles
Full title
Mixture Data for Training Cannot Ensure Out-of-distribution Generalization
Authors, Artists and Contributors
Author / Creator
Identifiers
Primary Identifiers
Record Identifier
TN_cdi_proquest_journals_2907599325
Permalink
https://devfeature-collection.sl.nsw.gov.au/record/TN_cdi_proquest_journals_2907599325
Other Identifiers
E-ISSN
2331-8422
