Online Label Shift: Optimal Dynamic Regret meets Practical Algorithms
This addresses the challenge of adapting machine learning models to dynamically changing label distributions in online settings, which is incremental as it builds on existing label shift adaptation methods.
The paper tackles the problem of online label shift, where class marginals change over time while class-conditionals stay the same, by developing algorithms for both supervised and unsupervised settings that adapt to drifting label distributions without prior knowledge. The result shows superior performance with 1-3% accuracy improvements in experiments across simulated and real-world scenarios.
This paper focuses on supervised and unsupervised online label shift, where the class marginals $Q(y)$ varies but the class-conditionals $Q(x|y)$ remain invariant. In the unsupervised setting, our goal is to adapt a learner, trained on some offline labeled data, to changing label distributions given unlabeled online data. In the supervised setting, we must both learn a classifier and adapt to the dynamically evolving class marginals given only labeled online data. We develop novel algorithms that reduce the adaptation problem to online regression and guarantee optimal dynamic regret without any prior knowledge of the extent of drift in the label distribution. Our solution is based on bootstrapping the estimates of \emph{online regression oracles} that track the drifting proportions. Experiments across numerous simulated and real-world online label shift scenarios demonstrate the superior performance of our proposed approaches, often achieving 1-3\% improvement in accuracy while being sample and computationally efficient. Code is publicly available at https://github.com/acmi-lab/OnlineLabelShift.