An Information-theoretic Approach to Distribution Shifts
This work addresses the challenge of safely deploying ML models in real-world scenarios where data shifts occur, such as across geographic locations or environments, but it is incremental as it builds on existing domain generalization and fair classification approaches.
The paper tackles the problem of data distribution shifts in machine learning deployment by proposing an information-theoretic perspective to analyze error sources and compare objectives from domain generalization and fair classification literature, concluding that model selection must consider observed data, correction factors, and data-generating structure.
Safely deploying machine learning models to the real world is often a challenging process. Models trained with data obtained from a specific geographic location tend to fail when queried with data obtained elsewhere, agents trained in a simulation can struggle to adapt when deployed in the real world or novel environments, and neural networks that are fit to a subset of the population might carry some selection bias into their decision process. In this work, we describe the problem of data shift from a novel information-theoretic perspective by (i) identifying and describing the different sources of error, (ii) comparing some of the most promising objectives explored in the recent domain generalization, and fair classification literature. From our theoretical analysis and empirical evaluation, we conclude that the model selection procedure needs to be guided by careful considerations regarding the observed data, the factors used for correction, and the structure of the data-generating process.