Toward Learning Robust and Invariant Representations with Alignment Regularization and Data Augmentation

Haohan Wang, Zeyi Huang, Xindi Wu, Eric P. Xing

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Data augmentation has been proven to be an effective technique for developing machine learning models that are robust to known classes of distributional shifts (e.g., rotations of images), and alignment regularization is a technique often used together with data augmentation to further help the model learn representations invariant to the shifts used to augment the data. In this paper, motivated by a proliferation of options of alignment regularizations, we seek to evaluate the performances of several popular design choices along the dimensions of robustness and invariance, for which we introduce a new test procedure. Our synthetic experiment results speak to the benefits of squared ℓ2 norm regularization. Further, we also formally analyze the behavior of alignment regularization to complement our empirical study under assumptions we consider realistic. Finally, we test this simple technique we identify (worst-case data augmentation with squared ℓ2 norm alignment regularization) and show that the benefits of this method outrun those of the specially designed methods. We also release a software package in both TensorFlow and PyTorch for users to use the method with a couple of lines at https://github.com/jyanln/AlignReg.
Original languageEnglish (US)
Title of host publicationKDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
PublisherAssociation for Computing Machinery
Pages1846-1856
ISBN (Print)9781450393850
DOIs
StatePublished - Aug 2022
Externally publishedYes

Keywords

  • robustness
  • trustworthy
  • data augmentation
  • machine learning

Fingerprint

Dive into the research topics of 'Toward Learning Robust and Invariant Representations with Alignment Regularization and Data Augmentation'. Together they form a unique fingerprint.

Cite this