Towards Principled Disentanglement for Domain Generalization
Authors:
Preprint
Abstract
A fundamental challenge for machine learning models is generalizing to out-of-distribution (OOD) data, in part due to spurious correlations. To tackle this challenge, we first formalize the OOD generalization problem as constrained optimization, called Disentanglement-constrained Domain Generalization (DDG). We relax this non-trivial constrained optimization to a tractable form with finite-dimensional parameterization and empirical approximation. Then a theoretical analysis of the extent to which the above transformations deviates from the original problem is provided. Based on the transformation, we propose a primal-dual algorithm for joint representation disentanglement and domain generalization. In contrast to traditional approaches based on domain adversarial training and domain labels, DDG jointly learns semantic and variation encoders for disentanglement, enabling flexible manipulation and augmentation on training data. DDG aims to learn intrinsic representations of semantic concepts that are invariant to nuisance factors and generalizable across different domains. Comprehensive experiments on popular benchmarks show that DDG can achieve competitive OOD performance and uncover interpretable salient structures within data.Keywords
All Related Versions
This publication has 21 references indexed in Scilit:
- Joint Discriminative and Generative Learning for Person Re-Identification2019
- Multimodal Unsupervised Image-to-Image Translation2018
- Deep Domain Generalization via Conditional Invariant Adversarial Networks2018
- Domain Generalization with Adversarial Feature Learning2018
- Image-to-Image Translation with Conditional Adversarial Networks2017
- Unbiased look at dataset bias2011
- A theory of learning from different domainsMachine Learning, 2009
- Reproducing Kernel Hilbert Spaces in Probability and Statistics2004
- Improving predictive inference under covariate shift by weighting the log-likelihood functionJournal of Statistical Planning and Inference, 2000
- Multilayer feedforward networks are universal approximatorsNeural Networks, 1989