terminological bleg: "structure learning"

[gusl]

When people say "structure learning", does this refer strictly to learning the structure of graphical models? Or more broadly learning distributions over structures of any kind?

Comments

[avani.livejournal.com]

It's almost always the structure of the underlying causality/interdependence relationships in your data. Once you have a structure, you can deduce a distribution across it. Does this help?

[gustavolacerda.livejournal.com]

i.e. "structure learning" = "I-map learning"?

I'm interested in learning structures with complex dependencies, such as probabilistic NL grammars. Does this count as "structure learning"?