arXiv：cs.LG· Guo Cheng, Changlong Lv, Jingyi Hou·· 3 小时前
CARE：面向长期时间序列预测的轻量级插件式门控修正与不确定性感知模块
CARE: A Lightweight Plug-in Gated Correction and Uncertainty-aware Module for Long-term Time Series Forecasting
AI 导读
研究者提出 CARE（Corrective branch with Aligned context and Relative-error Estimation），一个无需重构架构即可增强任意确定性预测器的轻量级插件。
正文
Abstract:Multivariate long-horizon forecasting is critical to electricity load scheduling and traffic flow management, and to financial risk control. Existing deterministic backbones output a single trajectory, masking heterogeneous prediction difficulty across horizons and channels and providing no localized reliability signal. We present CARE (Corrective branch with Aligned context and Relative-error Estimation), a lightweight plug-in that enhances any deterministic forecaster without architectural redesign. Operating in parallel with the base model, CARE resamples historical context to match the forecast horizon, learns residual correction patterns from this aligned history, and applies scale-aware bounded updates modulated by per-coordinate sigmoid risk gates. A multi-objective loss jointly optimizes forecast accuracy, residual tracking, risk alignment, and base-model anchoring. Across eight benchmarks with three representative backbones, CARE improves accuracy with marginal parameter and latency overhead. Its risk gates reliably identify high-error regions: on Weather, the highest-gate tertile exhibits nearly four times the error of the lowest-gate tertile, offering planners an interpretable per-step trust signal. Code is available at this https URL.
| Comments: | 15pages, 2figures |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.11165 [cs.LG] |
| (or arXiv:2610.11165v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.11165 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Guo Cheng [view email]
[v1]
Thu, 8 Oct 2026 03:23:39 UTC (115 KB)
来源：arXiv：cs.LG · arxiv.org