Robustly Train Normalizing Flows via KL Divergence Regularization
DOI:
https://doi.org/10.1609/aaai.v38i13.29426Keywords:
ML: Classification and Regression, ML: Deep Learning Algorithms, ML: Deep Neural Architectures and Foundation Models, ML: Learning with Manifolds, ML: Matrix & Tensor Methods, ML: Representation LearningAbstract
In this paper, we find that the training of Normalizing Flows (NFs) are easily affected by the outliers and a small number (or high dimensionality) of training samples. To solve this problem, we propose a Kullback–Leibler (KL) divergence regularization on the Jacobian matrix of NFs. We prove that such regularization is equivalent to adding a set of samples whose covariance matrix is the identity matrix to the training set. Thus, it reduces the negative influence of the outliers and the small sample number on the estimation of the covariance matrix, simultaneously. Therefore, our regularization makes the training of NFs robust. Ultimately, we evaluate the performance of NFs on out-of-distribution (OoD) detection tasks. The excellent results obtained demonstrate the effectiveness of the proposed regularization term. For example, with the help of the proposed regularization, the OoD detection score increases at most 30% compared with the one without the regularization.Downloads
Published
2024-03-24
How to Cite
Song, K., Solozabal, R., Li, H., Takáč, M., Ren, L., & Karray, F. (2024). Robustly Train Normalizing Flows via KL Divergence Regularization. Proceedings of the AAAI Conference on Artificial Intelligence, 38(13), 15047-15055. https://doi.org/10.1609/aaai.v38i13.29426
Issue
Section
AAAI Technical Track on Machine Learning IV