Towards Adaptive Time Series Foundation Models Against Distribution Shift
Keywords: Time series, Pretraining, Distribution Shifts, Foundation Model
Abstract: Foundation models have demonstrated remarkable success across diverse machine-learning domains through large-scale pretraining. However, their application to time series data poses challenges due to substantial mismatches in the distributions of pretraining datasets. In this paper, we tackle this issue by proposing a domain-aware adaptive normalization strategy within the Transformer architecture. Specifically, we replace the traditional LayerNorm with a prototype-guided dynamic normalization mechanism, where learned prototypes represent distinct data distributions, and sample-to-prototype similarity determines the appropriate normalization layer. This approach effectively captures the diverse characteristics of time series data, ensuring better alignment between pretrained representations and downstream tasks. Our method significantly improves fine-tuning performance, outperforming vanilla pretraining techniques and reducing the negative impact of distribution shifts. Extensive experiments on various real-world time series datasets demonstrate the efficacy of our approach, paving the way for more robust and generalizable time series foundation models.
Primary Area: learning on time series and dynamical systems
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 10509
Loading