研究人员开发了GRACE，一种用于在高维时间序列数据中发现因果关系的新颖方法。GRACE利用带有L0正则化的硬门控（Hard Concrete gates）来优化基于约束的发现，从而实现对因果边的稳健二元决策。与现有方法相比，该方法显著提高了F1分数和精确度，为复杂数据集提供了更快、更准确的解决方案。 AI
影响 该方法有望增进对气候和生物学等复杂系统的理解和预测。
排序理由 该集群包含一篇详细介绍因果发现新方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.23880v1 Announce Type: new Abstract: From climate teleconnections to gene regulation, modern time-series datasets encompass tens or hundreds of interacting variables, making causal discovery increasingly challenging. Constraint-based methods offer statistical rigor but…
From climate teleconnections to gene regulation, modern time-series datasets encompass tens or hundreds of interacting variables, making causal discovery increasingly challenging. Constraint-based methods offer statistical rigor but their nonlinear CI tests are infeasible at scal…