Learn to Expect the Unexpected: Probably Approximately Correct Domain Generalization

Abstract

Domain generalization is the problem of machine learning when the training data and the test data come from different data domains. We present a simple theoretical model of learning to generalize across domains in which there is a meta-distribution over data distributions, and those data distributions may even have different supports. In our model, the training data given to a learning algorithm consists of multiple datasets each from a single domain drawn in turn from the meta-distribution. We study this model in three different problem settings---a multi-domain Massart noise setting, a decision tree multi-dataset setting, and a feature selection setting, and find that computationally efficient, polynomial-sample domain generalization is possible in each. Experiments demonstrate that our feature selection algorithm indeed ignores spurious correlations and improves generalization.

References in corpus (8)

• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Domain Adaptation for Statistical Classifiers
• Domain Generalization via Invariant Feature Representation
• Domain Adaptation: Learning Bounds and Algorithms
• Domain Generalization via Model-Agnostic Learning of Semantic Features
• Risk bounds for statistical learning
• Learning Robust Representations by Projecting Superficial Statistics Out
• Multiple Source Adaptation and the Renyi Divergence