Community Structure and Link Prediction in Networks
Seminar Room 1, Newton Institute
AbstractWe investigate two questions of broad interest in the study of networks: the discovery of “communities” (regions of unusually high density within networks) and link prediction (judgements about which links in a network are mostly likely to be either false negatives or false positives). We approach both problems using likelihood-based fits to appropriate random graph models, including stochastic block models, edge-placement models, and hierarchical models. The overall conclusion is that these models appear to work well in general, giving useful answers to the questions of interest, but that there are a number of substantive issues, about which our understanding is currently poor, that prevent their wider application.
If it doesn't, something may have gone wrong with our embedded player.
We'll get it fixed as soon as possible.