Toward Learning Robust and Invariant Representations with Alignment Regularization and Data Augmentation
This work addresses the challenge of selecting effective regularization for robust machine learning, but it is incremental as it compares existing techniques rather than introducing a fundamentally new approach.
The paper evaluates various alignment regularization techniques for improving model robustness and invariance to distributional shifts, finding that squared l2 norm regularization with worst-case data augmentation outperforms specialized methods.
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.