Conference paper

Kleiter G. D., Jiroušek R.

Proceedings of the 8th Czech-Japan Seminar on Data Analysis and Decision Making under Uncertainty.

Oeconomica

Praha

2005

65-75

Prof. Radim Jiroušek, DrSc. (ÚTIA - Oddělení matematické teorie rozhodování)

Perfect Sequences: A Contribution to Structuring Conditional Independence Models [file]

The representation of conditional independence models by perfect sequences provides an alternative to Bayesian networks and essential graphs. The paper discusses properties of perfect sequences that are relevant with respect to different structures of conditional independence models. Boundary variables (related to terminal nodes in a directed graph representation) are used to find the number of labeled and unlabeled models and to enumerate parts of the model space. Structuring principles are further applied to the evaluation of whole conditional independence models in learning models from data.