arXiv cs.AI· Songlin Du, Ling Luo, Uwe Aickelin·· 12 小时前AI 评分22
DDCI：面向时间序列预测的动态双反馈保形推理
Dynamic Dual-Feedback Conformal Inference for Time Series Forecasting
AI 导读
论文提出动态双反馈保形推理（DDCI），在原有仅依赖实际反馈（真实标签是否落入区间）的在线保形方法上，加入来自 oracle 保形方法的经验预测阈值作为估计反馈，以抑制区间的过度扩张或收缩。DDCI 在序列场景下实现了更稳定、更窄的预测区间，同时在实际中保持接近名义水平的覆盖率。该工作已被 NeurIPS 2026 接收。
正文 · 原文
Abstract:Quantifying uncertainty in time series forecasting is particularly demanding because sequential data exhibit temporal dependence and are prone to distributional changes. Conformal inference has emerged as a powerful uncertainty quantification approach through the construction of prediction sets. Recent advances have introduced online conformal methods that adaptively adjust prediction thresholds through feedback mechanisms. However, the existing feedback mechanism typically relies solely on miscoverage indicators (actual feedback), whether the true label falls within the interval at each time step, while overlooking the empirical prediction threshold (estimated feedback) that is derived from the oracle conformal method. In this paper, we propose Dynamic Dual-feedback Conformal Inference (DDCI), which incorporates a dual-feedback mechanism consisting of actual feedback and estimated feedback. The former drives the primary adjustment of the intervals based on true observations, while the latter dampens excessive expansions or contractions by leveraging empirical thresholds from conformal inference during updates. By balancing these two signals, DDCI achieves more stable and narrower prediction intervals in sequential settings while maintaining near-nominal coverage in practice.
| Comments: | Accepted by NeurIPS 2026 |
| Subjects: | Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.26987 [stat.ML] |
| (or arXiv:2610.26987v1 [stat.ML] for this version) | |
| https://doi.org/10.48550/arXiv.2610.26987 arXiv-issued DOI via DataCite |
Submission history
From: Songlin Du [view email]
[v1]
Mon, 19 Oct 2026 05:28:22 UTC (15,329 KB)
来源：arXiv cs.AI · arxiv.org