Date(s) - 11/04/2013
3:00 am - 4:30 am
Recent work in computational social science has made heavy use of statistical models for characterizing the probability of network configurations and their change over time. We will review several statistical network models, including: (a) the exponential random graph model, where the probability of the network is a log-linear function of feature counts; (b) the stochastic block-model, where edge likelihoods are governed by membership in latent communities; (c) the mixed-membership stochastic block-model — a network analogue of latent Dirichlet allocation — where nodes participate in multiple latent communities.
This will be an informal discussion in the style of a reading group, and will center on the survey of Goldenberg et al (2009) (http://arxiv.org/abs/0912.5410/), covering chapters 1, 2, 5, 6, and parts of chapter 3 (to be announced). ***Participants are expected to have read this material ahead of time to facilitate an interactive discussion.***