arXiv：cs.LG· Hao Wang, Licheng Pan, Zhichao Chen, Xu Chen, Qingyang Dai, Lei Wang, Haoxuan Li, Zhouchen Lin·· 7 小时前AI 评分41
Time-o1：时间序列预测需要变换后的标签对齐
Time-o1: Time-Series Forecasting Needs Transformed Label Alignment
AI 导读
针对时间序列预测中时序均方误差忽略标签自相关、任务数量过多的问题，研究者提出变换增强损失函数 Time-o1，将标签序列变换为去相关且显著性分明的分量，让模型对齐其中最重要的部分，从而缓解标签自相关并减少任务量。实验显示 Time-o1 达到 SOTA 性能，并兼容多种预测模型，代码已公开，论文被 NeurIPS 2025 接收为 poster。
正文
Abstract:Training time-series forecasting models poses unique challenges in loss function design. Most existing approaches adopt temporal mean squared error, but this study reveals two critical limitations: (1) it ignores the presence of label autocorrelation, which biases it from the true label sequence likelihood; (2) it involves excessive number of tasks, which complicates optimization, especially for long-term forecasting. To address these issues, we introduce Time-o1, a transform-enhanced loss function for time-series forecasting. The central idea is to transform the label sequence into decorrelated components with discriminated significance. Models are then trained to align the most significant components, thereby effectively mitigating label autocorrelation and reducing task amount. Experiments demonstrate that Time-o1 achieves state-of-the-art performance and is compatible with various forecast models. Code is available at this https URL.
| Comments: | Accepted as poster in NeurIPS 2025 |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Systems and Control (eess.SY) |
| Cite as: | arXiv:2505.17847 [cs.LG] |
| (or arXiv:2505.17847v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2505.17847 arXiv-issued DOI via DataCite |
|
| Journal reference: | NeurIPS 2025 |
Submission history
From: Hao Wang [view email]
[v1]
Fri, 23 May 2025 13:00:35 UTC (2,074 KB)
[v2]
Thu, 2 Oct 2025 13:18:08 UTC (2,080 KB)
[v3]
Tue, 6 Oct 2026 12:21:17 UTC (2,099 KB)
来源：arXiv：cs.LG · arxiv.org