Learning Representations Theory
12. Learning Representations Theory#
In the main text, you learned many approaches for representing networks. These methods, while in general rather principled, also have substantial mathematical foundations as to why they are sensible. In particular, when you make assumptions about the network sample that you have, you can in general prove that the approaches we described in the main text provide reasonable algorithmic approaches. We outline more in-depth discussions on the sections here:
Maximum Likelihood Estimate Theory discusses maximum likelihood approaches to ER and SBM networks, and
Spectral Method Theory discusses consistent estimators of latent position matrices and joint matrices.