On deducing conditional independence from d-separation in causal graphs with feedback

Authors
Citation
Rm. Neal, On deducing conditional independence from d-separation in causal graphs with feedback, J ARTIF I R, 12, 2000, pp. 87-91
Citations number
3
Categorie Soggetti
AI Robotics and Automatic Control
Journal title
JOURNAL OF ARTIFICIAL INTELLIGENCE RESEARCH
ISSN journal
10769757 → ACNP
Volume
12
Year of publication
2000
Pages
87 - 91
Database
ISI
SICI code
1076-9757(2000)12:<87:ODCIFD>2.0.ZU;2-J
Abstract
Pearl and Dechter (1996) claimed that the d-separation criterion for condit ional independence in acyclic causal networks also applies to networks of d iscrete variables that have feedback cycles, provided that the variables of the system are uniquely determined by the random disturbances. I show by e xample that this is not true in general. Some condition stronger than uniqu eness is needed, such as the existence of a causal dynamics guaranteed to l ead to the unique solution.