A Unifying Causal Framework for Analyzing Dataset Shift-stable Learning Algorithms
This work addresses the challenge of external validity for machine learning practitioners by providing a theoretical foundation to analyze and design stable learning algorithms, though it is incremental in building on existing causal graphical methods.
The authors tackled the problem of dataset shift in prediction models by developing a unifying causal framework that expresses various shift types through graphical operators, establishing conditions for minimax optimal performance and deriving new algorithms, with empirical results showing a tradeoff between minimax and average performance.
Recent interest in the external validity of prediction models (i.e., the problem of different train and test distributions, known as dataset shift) has produced many methods for finding predictive distributions that are invariant to dataset shifts and can be used for prediction in new, unseen environments. However, these methods consider different types of shifts and have been developed under disparate frameworks, making it difficult to theoretically analyze how solutions differ with respect to stability and accuracy. Taking a causal graphical view, we use a flexible graphical representation to express various types of dataset shifts. Given a known graph of the data generating process, we show that all invariant distributions correspond to a causal hierarchy of graphical operators which disable the edges in the graph that are responsible for the shifts. The hierarchy provides a common theoretical underpinning for understanding when and how stability to shifts can be achieved, and in what ways stable distributions can differ. We use it to establish conditions for minimax optimal performance across environments, and derive new algorithms that find optimal stable distributions. Using this new perspective, we empirically demonstrate that that there is a tradeoff between minimax and average performance.