arXiv：cs.LG· Hao Wang, Licheng Pan, Zhichao Chen, Degui Yang, Sen Zhang, Yifei Yang, Xinggao Liu, Haoxuan Li, Dacheng Tao·· 7 小时前AI 评分34
FreDF：在频域中学习预测的时间序列模型
FreDF: Learning to Forecast in the Frequency Domain
AI 导读
针对现有 Direct Forecast（DF）范式忽略未来标签自相关、导致学习目标有偏的问题，研究者提出频域增强直接预测方法 FreDF，通过在频域中学习预测来缓解标签自相关并降低估计偏差。实验显示 FreDF 显著优于现有 SOTA 方法，并兼容多种预测模型，代码已开源，该工作已被 ICLR 2025 接收。
正文
Abstract:Time series modeling presents unique challenges due to autocorrelation in both historical data and future sequences. While current research predominantly addresses autocorrelation within historical data, the correlations among future labels are often overlooked. Specifically, modern forecasting models primarily adhere to the Direct Forecast (DF) paradigm, generating multi-step forecasts independently and disregarding label autocorrelation over time. In this work, we demonstrate that the learning objective of DF is biased in the presence of label autocorrelation. To address this issue, we propose the Frequency-enhanced Direct Forecast (FreDF), which mitigates label autocorrelation by learning to forecast in the frequency domain, thereby reducing estimation bias. Our experiments show that FreDF significantly outperforms existing state-of-the-art methods and is compatible with a variety of forecast models. Code is available at this https URL.
| Comments: | Accepted by ICLR 2025 |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Applications (stat.AP); Machine Learning (stat.ML) |
| Cite as: | arXiv:2402.02399 [cs.LG] |
| (or arXiv:2402.02399v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2402.02399 arXiv-issued DOI via DataCite |
Submission history
From: Hao Wang [view email]
[v1]
Sun, 4 Feb 2024 08:23:41 UTC (5,263 KB)
[v2]
Tue, 6 May 2025 06:56:48 UTC (6,245 KB)
[v3]
Tue, 6 Oct 2026 12:07:27 UTC (4,413 KB)
来源：arXiv：cs.LG · arxiv.org