Learning and Predicting Dynamic Network Behavior with Graphical Multiagent Models

Factored models of multiagent systems address the complexity of joint behavior by exploiting locality in agent interactions. History-dependent graphical multiagent models (hGMMs) further capture dynamics by conditioning behavior on history. The hGMM framework also brings new elements of strategic reasoning and more expressive powers to modeling information diffusion over networks. We propose a greedy algorithm for learning hGMMs from time-series data, inducing both graphical structure and parameters. To evaluate this learning method, we employ human-subject experiment data for a voting consensus scenario, where agents on a network attempt to reach a unanimous vote. We empirically show that the learned hGMMs directly expressing joint behavior outperform alternatives in predicting dynamic voting behavior.