BriefGPT.xyz
Jun, 2011
从有反馈的因果图中用d-分离推导条件独立性（研究笔记）
On Deducing Conditional Independence from d-Separation in Causal Graphs with Feedback (Research Note)
HTML
PDF
R. M. Neal
TL;DR
本文旨在反驳Pearl和Dechter（1996）提出的基于d-分离准则的有向无环因果网络的条件独立性也适用于存在反馈循环的离散变量网络（前提是系统的变量由随机干扰确定），我通过示例说明通常情况下这并不成立，需要更强的条件，如存在可以保证唯一解的因果动态。
Abstract
Pearl and Dechter (1996) claimed that the d-separation criterion for
conditional independence
in acyclic
causal networks
also applies to networks of discrete variables that have
→