论文

Conformal Expert Pools:零训练概率时间序列预测方法 CEP 发布

Conformal Expert Pools: Training-Free Probabilistic Time-Series Forecasting Beyond Seasonality

精选理由

不用训练任何模型,只用历史误差构建预测分布,就在 232 个数据集上压过 Amazon 的 DeepNPTS,区间窄 40%、快 271 倍。做时间序列预测的工程师值得把 CEP 的做法直接拿来当基线,可能省掉一整套深度学习管线。

一篇 arXiv 论文提出 Conformal Expert Pools(CEP),这是 Conformal Seasonal Pools(CSP)的后继方法。CEP 不训练任何模型、不跨序列学习参数,而是用一组简单预测器组成的池,基于它们的历史预测误差构建预测分布,并针对待预测序列自适应调整池和误差。

正文

在 CSP 基准的 232 个数据集、24,502 个预测窗口上,CEP 按 CRPS 在 78% 的数据集上优于 CSP,按 MSIS 在 84% 上更优,按 Winkler 分数在 83% 上更优;典型数据集上 CRPS 降低 14%、MSIS 降低 23%(配对 Wilcoxon 检验 p < 1e-22),95% 区间覆盖率从 90.1% 升至 92.2%。与 Amazon 的 DeepNPTS、NPTS、SeasonalNPTS 对比,CEP 分别在 88%、89%、83% 的数据集上 CRPS 更低。改进在开发集之外的数据集家族和 36 个额外 fev-bench 文件上依然成立。

论文还检验了在线集成预测器 Laplace "几乎击败一切"的说法:在 228 个数据集的多步预测上,CSP 按 CRPS 在 60% 的数据集上胜过 Laplace,Laplace 在七种方法中排第六,区间比 CEP 宽 69%,冷启动耗时约为 CSP 的 7,700 倍;CEP 按 CRPS 在 82% 的数据集上优于 Laplace,区间窄约 40%,速度快 271 倍。

原文 · arXiv cs.AI

Conformal Expert Pools: Training-Free Probabilistic Time-Series Forecasting Beyond Seasonality

Conformal Seasonal Pools (CSP) showed that a forecaster with no training at all, built only from past values at the same point in the season and from past forecast errors, significantly outperformed DeepNPTS, Amazon's learned non-parametric forecaster. We propose Conformal Expert Pools (CEP), a training-free successor. CEP replaces the single seasonal-naive forecast by a pool of simple forecasters, builds its predictive distribution from their past forecast errors, and adapts both the pool and the errors to the series being forecast. No model is trained and no parameter is learned across series. On the CSP benchmark of 232 datasets and 24,502 forecast windows, CEP is better than CSP on 78% of datasets by CRPS, on 84% by MSIS and on 83% by Winkler score; on a typical dataset its CRPS is 14% lower and its MSIS 23% lower (both p < 1e-22, paired Wilcoxon test), and coverage of 95% intervals rises from 90.1% to 92.2%. CEP has a lower CRPS than DeepNPTS on 88% of datasets, than NPTS on 89% and than SeasonalNPTS on 83%. The improvement persists on dataset families outside the development set and on 36 additional fev-bench files. We also test the published claims that Laplace, an online ensemble forecaster, "beats (almost) everything" and "beats" CSP. On 228 datasets at multi-step horizons the claims do not hold: CSP outperforms Laplace on 60% of datasets by CRPS and on 64% by interval score, and Laplace ranks sixth of seven methods, with the widest intervals of all (69% wider than CEP's). Its point forecast is biased low and on 54% of datasets is less accurate than seasonal naive, and from a cold start it takes about 7,700 times longer than CSP. CEP is better than Laplace on 82% of datasets by CRPS and on 80% by MSIS, with intervals about 40% narrower, and runs 271 times faster.

产业坐标与信源核验

本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:Conformal Prediction、时间序列预测、DeepNPTS、CRPS,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明:本文原始线索采编自「arXiv cs.AI」(@Valery Manokhin),经 AITOP 多维评分管线结构化提炼。