Machine Learning Work Shop - Graphical Event Models for Temporal Event Streams
7Machine Learning Work Shop - Session 3 - Asela Gunawardana - 'Graphical Event Models for Temporal Event Streams' Many phenomena can be described as streams of diverse events in time. Examples include the stream of actions a user makes when navigating and searching on the web or when using their mobile phone, and the stream of system events in a datacenter. Such streams can be modeled probabilistically as marked point process in time. Modeling the dependencies between events in time is important for understanding these processes , for forecasting their future evolution, and for planning in the face of the uncertainty regarding their future evolution. We present our recent work on building models of these dependencies and developing algorithms for forecasting and planning in these settings, as well as results in applying these methods to a number of real world applications.