Support and Invertibility in Domain-Invariant Representations

Abstract

Learning domain-invariant representations has become a popular approach to unsupervised domain adaptation and is often justified by invoking a particular suite of theoretical results. We argue that there are two significant flaws in such arguments. First, the results in question hold only for a fixed representation and do not account for information lost in non-invertible transformations. Second, domain invariance is often a far too strict requirement and does not always lead to consistent estimation, even under strong and favorable assumptions. In this work, we give generalization bounds for unsupervised domain adaptation that hold for any representation function by acknowledging the cost of non-invertibility. In addition, we show that penalizing distance between densities is often wasteful and propose a bound based on measuring the extent to which the support of the source domain covers the target domain. We perform experiments on well-known benchmarks that illustrate the short-comings of current standard practice.

References in corpus (3)

• Domain Separation Networks
• Domain Generalization via Invariant Feature Representation
• On Learning Invariant Representation for Domain Adaptation

Cited by in corpus (8)

• On Learning Invariant Representation for Domain Adaptation
• In Search of Lost Domain Generalization
• Domain Aggregation Networks for Multi-Source Domain Adaptation
• Continuous Transfer Learning with Label-informed Distribution Alignment
• Learning to Match Distributions for Domain Adaptation
• On Localized Discrepancy for Domain Adaptation
• Regret Minimization for Causal Inference on Large Treatment Space
• The Role of Embedding Complexity in Domain-invariant Representations