Towards Principled Test-Time Adaptation for Time Series Forecasting

TL;DR

This paper introduces a principled test-time adaptation protocol for time series forecasting that leverages frequency domain calibration, improving performance with fewer parameters under distribution shifts.

Contribution

It proposes a unified, frequency-aware calibration method for TSF-TTA that is more principled, lightweight, and effective across diverse datasets and forecasting horizons.

Findings

FAC achieves competitive performance across datasets.

Prediction corrections show limited spectral modifications.

Fewer trainable parameters than existing adapters.

Abstract

Test-time adaptation (TTA) has recently emerged as a promising approach for improving time series forecasting (TSF) under distribution shift. Existing TSF-TTA methods differ in how they utilize revealed targets, yet the resulting adaptation protocols remain heterogeneous and lack a clearly unified formulation. To address this issue, we revisit TSF-TTA from the perspective of protocol cleanliness and propose an adaptation protocol based solely on matured ground truth, yielding a more principled setting for adaptation. Under this protocol, we further diagnose existing adapters in the frequency domain and find that their prediction corrections often exhibit limited and weakly structured spectral modifications. Motivated by this diagnosis, we propose Frequency-Aware Calibration (FAC), a lightweight calibration method that directly parameterizes prediction corrections in the frequency domain. Across diverse datasets, forecasting horizons, and source forecasters, FAC achieves competitive and consistent performance while requiring substantially fewer trainable parameters than the compared TSF-TTA adapters.