Model Assessment and Selection under Temporal Distribution Shift

TL;DR

This paper proposes an adaptive method for model assessment and selection under temporal distribution shifts by synthesizing datasets, estimating generalization errors, and using a tournament approach, with strong theoretical and experimental support.

Contribution

It introduces a novel adaptive rolling window approach for error estimation and a tournament-based model selection method tailored for non-stationary environments.

Findings

The method effectively estimates generalization error under distribution shift.

The tournament approach achieves near-optimal model selection.

The approach is supported by theoretical analysis and numerical experiments.

Abstract

We investigate model assessment and selection in a changing environment, by synthesizing datasets from both the current time period and historical epochs. To tackle unknown and potentially arbitrary temporal distribution shift, we develop an adaptive rolling window approach to estimate the generalization error of a given model. This strategy also facilitates the comparison between any two candidate models by estimating the difference of their generalization errors. We further integrate pairwise comparisons into a single-elimination tournament, achieving near-optimal model selection from a collection of candidates. Theoretical analyses and numerical experiments demonstrate the adaptivity of our proposed methods to the non-stationarity in data.