Showing 1–1 of 1 results for author: Roizer, I

Learning Link-Probabilities in Causal Trees

Authors: Igor Roizer, Judea Pearl

Abstract: A learning algorithm is presented which given the structure of a causal tree, will estimate its link probabilities by sequential measurements on the leaves only. Internal nodes of the tree represent conceptual (hidden) variables inaccessible to observation. The method described is incremental, local, efficient, and remains robust to measurement imprecisions. A learning algorithm is presented which given the structure of a causal tree, will estimate its link probabilities by sequential measurements on the leaves only. Internal nodes of the tree represent conceptual (hidden) variables inaccessible to observation. The method described is incremental, local, efficient, and remains robust to measurement imprecisions. △ Less

Submitted 27 March, 2013; originally announced April 2013.

Comments: Appears in Proceedings of the Second Conference on Uncertainty in Artificial Intelligence (UAI1986)

Report number: UAI-P-1986-PG-211-214