V-structure

From Machinelearning
Revision as of 00:36, 22 September 2018 by IssaRice (talk | contribs)

In graphical models, a v-structure is a set of three nodes in a graph such that there are two converging arrows and the tails of the arrows are not connected by an arrow.[1][2] For example, X→Z←Y is a v-structure because the arrows converge at Z and the nodes X and Y are not directly connected.

The middle node in a v-structure is called a collider. (?)

Some authors seem to distinguish between a v-structure and an immoral v-structure. In this case, "v-structure" seems to only mean that there is a collision, and the "immoral" means the parents are not directly connected.[3][4]

Terminology

why is this thing called a v-structure? because it looks like a letter V if the two parents are drawn above the common child?

References

  1. ↑ Judea Pearl. Causality. p. 19.
  2. ↑ "Bayesian networks". Retrieved September 21, 2018.
  3. ↑ Su-In Lee. "Bayesian Network Representation". University of Washington CSE 515, Statistical Methods, Spring 2011. p. 14 (slide 27). March 30, 2011. Retrieved September 21, 2018.
  4. ↑ Dave Essary. "Bayesian Networks Representation". p. 15. Retrieved September 21, 2018.