Jump to content

Talk:Junction tree algorithm

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Wiki Education Foundation-supported course assignment

[edit]

This article was the subject of a Wiki Education Foundation-supported course assignment, between 27 August 2018 and 15 December 2018. Further details are available on the course page. Student editor(s): Kboback. Peer reviewers: MSan123, Cubby1999.

Above undated message substituted from Template:Dashboard.wikiedu.org assignment by PrimeBOT (talk) 01:32, 17 January 2022 (UTC)[reply]

Hugin

[edit]

I added several punctuations to the Hugin section as well as several hyperlinks.Cubby1999 (talk) 19:19, 31 October 2018 (UTC)[reply]

Spanning tree?

[edit]

The page says: "Construct a junction tree from this (form a maximal spanning tree)". What is the relationship between a junction tree and maximal spanning tree? A5 (talk) 12:58, 9 September 2008 (UTC)[reply]

The phrase in the parentheses implies the junction tree formation is similar to forming a spanning tree on the original graph, but that's misleading: to form a junction tree you first construct a hypergraph according to cliques in the triangulated graph, and then you construct a spanning tree on those supernodes. To answer your question in short, the junction tree is a maximal spanning tree, but not on the original graph (instead, it's on a graph constructed from the original graph). 128.31.34.229 (talk) 15:12, 2 December 2008 (UTC)[reply]

Jordan

[edit]

Contact Michael I Jordan- he pretty much invented the junction tree and will be able to provide the background information for this page. His book, which covers it in Chapter 17, is due to be published in the next year or two. —Preceding unsigned comment added by 79.68.59.167 (talk) 08:24, 23 October 2008 (UTC)[reply]

But didn't Lauritzen and Spiefelhalter invent it in their 1988 paper? Also, where are the references for Hugin and Shafer-Shenoy? 115.129.16.201 (talk) 14:45, 27 June 2009 (UTC) Dmitry Kamenetsky[reply]

Lauritzen is the earliest use of tree decomposition for inference that I can see, but essentially the same algorithm has been reinvented many times and used earlier without being specifically applied to inference, for instance in Arnborg's 1985 paper "Efficient algorithms for combinatorial problems on graphs with bounded decomposability". Boedlander has an overview Yaroslav Bulatov (talk) 19:11, 7 November 2010 (UTC)[reply]