An implementation of the iterative proportional fitting procedure by propagation trees

被引：17
|
作者：
Badsberg, JH
Malvestuto, FM
机构：
[1] Univ Roma La Sapienza, Dipartimento Sci Informaz, I-00198 Rome, Italy
[2] Res Ctr Foulum, Dept Agr Syst, Biometry Res Unit, DK-8830 Tjele, Denmark
来源：
COMPUTATIONAL STATISTICS & DATA ANALYSIS | 2001年 / 37卷 / 03期
关键词：
acyclic hypergraph; iterative proportional fitting procedure; Markov extension; probabilistic database; propagation tree;
D O I：
10.1016/S0167-9473(01)00013-5
中图分类号：
TP39 [计算机的应用];
学科分类号：
081203 ; 0835 ;
摘要：
The space-saving implementation of the iterative proportional fitting procedure proposed by Jirousek and Preucil (Comput. Statist. Data Anal. 19 (1995) 177) on this journal can be improved by applying the tree-computation techniques designed for Markov networks. The optimisation problem raised by the use of Markovian propagation trees is solved. Next, an even better implementation is obtained using certain trees, here introduced and called fast propagation trees, which are obtained by "simplifying" optimal Markovian propagation trees. (C) 2001 Elsevier Science B.V. All rights reserved.
引用
收藏
页码：297 / 322
页数：26
相关论文
共 50 条
12345