search

open-connectome-classes / StatConn-Spring-2015-Info

introductory material
18 stars 4 forks source link

Comparing different block models using ARI #247

Open dlee138 opened 9 years ago

dlee138 commented 9 years ago

If I want to show that one model (i.e. SBM) does a better job than another in a clustering problem, would an adjusted rand index (ARI) be appropriate for this and how would I implement it? I've seen some cases where the ARI is bound by [0,1] and other cases its [-1,1] and I'm confused as to whether it is a standard to implement negative values where the model is even worse than chance. Also, any other tools I could use other than the ARI for comparing different SBMs?

mblohr commented 9 years ago

See issue #90: Permutations vs. Adjusted Rand Index