Battling the Non-stationarity in Time Series Forecasting via Test-time Adaptation
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915095636869120 |
|---|---|
| author | Kim, HyunGi Kim, Siwon Mok, Jisoo Yoon, Sungroh |
| author_facet | Kim, HyunGi Kim, Siwon Mok, Jisoo Yoon, Sungroh |
| contents | Deep Neural Networks have spearheaded remarkable advancements in time series forecasting (TSF), one of the major tasks in time series modeling. Nonetheless, the non-stationarity of time series undermines the reliability of pre-trained source time series forecasters in mission-critical deployment settings. In this study, we introduce a pioneering test-time adaptation framework tailored for TSF (TSF-TTA). TAFAS, the proposed approach to TSF-TTA, flexibly adapts source forecasters to continuously shifting test distributions while preserving the core semantic information learned during pre-training. The novel utilization of partially-observed ground truth and gated calibration module enables proactive, robust, and model-agnostic adaptation of source forecasters. Experiments on diverse benchmark datasets and cutting-edge architectures demonstrate the efficacy and generality of TAFAS, especially in long-term forecasting scenarios that suffer from significant distribution shifts. The code is available at https://github.com/kimanki/TAFAS. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_04970 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Battling the Non-stationarity in Time Series Forecasting via Test-time Adaptation Kim, HyunGi Kim, Siwon Mok, Jisoo Yoon, Sungroh Machine Learning Artificial Intelligence Deep Neural Networks have spearheaded remarkable advancements in time series forecasting (TSF), one of the major tasks in time series modeling. Nonetheless, the non-stationarity of time series undermines the reliability of pre-trained source time series forecasters in mission-critical deployment settings. In this study, we introduce a pioneering test-time adaptation framework tailored for TSF (TSF-TTA). TAFAS, the proposed approach to TSF-TTA, flexibly adapts source forecasters to continuously shifting test distributions while preserving the core semantic information learned during pre-training. The novel utilization of partially-observed ground truth and gated calibration module enables proactive, robust, and model-agnostic adaptation of source forecasters. Experiments on diverse benchmark datasets and cutting-edge architectures demonstrate the efficacy and generality of TAFAS, especially in long-term forecasting scenarios that suffer from significant distribution shifts. The code is available at https://github.com/kimanki/TAFAS. |
| title | Battling the Non-stationarity in Time Series Forecasting via Test-time Adaptation |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2501.04970 |