HyperTime: Hyperparameter Optimization for Combating Temporal Distribution Shifts
This work addresses the challenge of maintaining model performance over time for practitioners dealing with evolving data distributions, though it is incremental in applying robust optimization principles to hyperparameter tuning.
The authors tackled the problem of temporal distribution shifts in unseen test data by proposing HyperTime, a hyperparameter optimization method that prioritizes worst-case validation loss over chronological sets, achieving robust predictive performance across multiple machine learning tasks.
In this work, we propose a hyperparameter optimization method named \emph{HyperTime} to find hyperparameters robust to potential temporal distribution shifts in the unseen test data. Our work is motivated by an important observation that it is, in many cases, possible to achieve temporally robust predictive performance via hyperparameter optimization. Based on this observation, we leverage the `worst-case-oriented' philosophy from the robust optimization literature to help find such robust hyperparameter configurations. HyperTime imposes a lexicographic priority order on average validation loss and worst-case validation loss over chronological validation sets. We perform a theoretical analysis on the upper bound of the expected test loss, which reveals the unique advantages of our approach. We also demonstrate the strong empirical performance of the proposed method on multiple machine learning tasks with temporal distribution shifts.