BriefGPT.xyz
Mar, 2019
使用随机复杂度测试离散数据的条件独立性
Testing Conditional Independence on Discrete Data using Stochastic Complexity
HTML
PDF
Alexander Marx, Jilles Vreeken
TL;DR
本文提出了一种基于算法独立性的、使用随机复杂性解决离散数据条件互信息估计问题的测试方法SCI，此方法可以在有限的样本上找到合理的CMI阈值。实验证明SCI比常规测试具有更低的II类错误和更高的召回率，可应用于因果发现算法中。
Abstract
Testing for
conditional independence
is a core aspect of constraint-based
causal discovery
. Although commonly used tests are perfect in theory, they often fail to reject independence in practice, especially when
→