Learning Domain Invariant Representations by Joint Wasserstein Distance Minimization

Abstract

Domain shifts in the training data are common in practical applications of machine learning; they occur for instance when the data is coming from different sources. Ideally, a ML model should work well independently of these shifts, for example, by learning a domain-invariant representation. However, common ML losses do not give strong guarantees on how consistently the ML model performs for different domains, in particular, whether the model performs well on a domain at the expense of its performance on another domain. In this paper, we build new theoretical foundations for this problem, by contributing a set of mathematical relations between classical losses for supervised ML and the Wasserstein distance in joint space (i.e. representation and output space). We show that classification or regression losses, when combined with a GAN-type discriminator between domains, form an upper-bound to the true Wasserstein distance between domains. This implies a more invariant representation and also more stable prediction performance across domains. Theoretical results are corroborated empirically on several image datasets. Our proposed approach systematically produces the highest minimum classification accuracy across domains, and the most invariant representation.

23 pages + supplement

References in corpus (12)

• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Improved Regularization of Convolutional Neural Networks with Cutout
• Domain Generalization: A Survey
• Domain Generalization via Invariant Feature Representation
• Domain Generalization via Model-Agnostic Learning of Semantic Features
• WILDS: A Benchmark of in-the-Wild Distribution Shifts
• Domain Generalization for Medical Imaging Classification with Linear-Dependency Regularization
• Model Patching: Closing the Subgroup Performance Gap with Data Augmentation
• Domain Generalization with MixStyle
• Semi-Supervised Domain Adaptation with Non-Parametric Copulas
• Domain Generalization via Semi-supervised Meta Learning
• Bures Joint Distribution Alignment with Dynamic Margin for Unsupervised Domain Adaptation

Cited by in corpus (1)

• Insightful analysis of historical sources at scales beyond human capabilities using unsupervised Machine Learning and XAI