arXiv cs.AI· Kexuan Zhang, Paolo Frazzetto, Xiaobei Zou, Gustau Camps-Valls, Yang Tang·· 3 小时前AI 评分34
CoLaX：面向时间序列分类的上下文感知局部解释框架
CoLaX: Context-Aware Local Explanations for Time Series Classification
AI 导读
研究者提出 CoLaX，一种上下文感知的时间序列分类事后解释框架，基于 shapelet 驱动的 Shapley 归因，并用全序列学习表示对局部模式的关联度进行条件化。在相同 motif 随上下文改变重要性的合成基准上，CoLaX 的 AUPRC 比最强基线平均提升 30%，并在真实 ECG 数据上取得最高 AUPRC。消融实验显示增益来自依赖输入的条件化而非额外可学习权重。
正文 · 原文
Abstract:Post-hoc explanations for time series classifiers usually score local patterns, such as subsequences or shapelets, without regard to the rest of the sequence. Yet the same motif can carry different predictive meaning under different global temporal conditions, such as trends or regimes. We propose CoLaX, a context-aware post-hoc explanation framework that builds on shapelet-driven Shapley attribution and conditions local-pattern relevance on a learned representation of the full sequence. Global context acts at two levels: it produces sample-specific prototype weights and a temporally resolved signal that reweights the local activation map. The resulting context-conditioned segments are the players of a Shapley attribution game whose values come only from the frozen black-box classifier. On synthetic benchmarks where identical motifs change relevance across contexts, CoLaX improves AUPRC over the strongest baseline by 30\% on average. On real-world data, it achieves the highest AUPRC on ECG. Additional ablations indicate that the gains come from input-dependent conditioning rather than from additional learnable weights. CoLaX is therefore most useful when recurring local motifs change relevance across sequence-level contexts.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.15067 [cs.LG] |
| (or arXiv:2611.15067v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.15067 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Kexuan Zhang [view email]
[v1]
Wed, 11 Nov 2026 05:36:26 UTC (1,841 KB)
来源：arXiv cs.AI · arxiv.org