Junction Tree

A junction tree is a method of organizing and understanding data, using a graph which branches off, like a tree, and where each branch holds a clustered subset of data that relates to a certain variable. It's useful because the junctions — where the subsets of data meet — show where they overlap, or where there are similarities in the data.

Junction tree algorithms, meanwhile, are used in machine learning, where datasets can be massive and where the goal is to calculate probabilities, such as the likelihood of a person buying a product or their having an accident while crossing a busy road. Junction trees are sometimes referred to as cinque trees, but these are not necessarily or always the same thing. (za/fs)

 

Sources:

A Short Course on Graphical Models (Mark Paskin, Stanford) https://ai.stanford.edu/~paskin/gm-short-course/lec3.pdf (accessed October 18, 2023)

The Junction Tree Algorithm (Berkeley) https://people.eecs.berkeley.edu/~jordan/courses/281A-fall04/lectures/lec-11-16.pdf (accessed October 18, 2023)

The Junction Tree Algorithm (Chris Williams, University of Edinburgh) https://www.inf.ed.ac.uk/teaching/courses/pmr/slides/jta-2x2.pdf (accessed October 18, 2023)

Read the rest of DW's AI Lexicon:

A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z

We're keen to hear your feedback. Suggest an entry by sending us a comment. And let us know if you feel we have missed something, got it wrong, and tell us whether our AI Lexicon has helped you understand the technology better.

Written and edited by: Zulfikar Abbany (za), Fred Schwaller (fs)