Many domains in the real world are richly structured, containing a diverse set of objects, related to each other in a variety of ways. For example, a living cell contains a rich network of interacting genes, that come together to perform key functions. A robot scan of a physical environment contains diverse objects such as people, vehicles, trees, or buildings, each of which might itself be a structured object. And a website contains a set of interlinked webpages, representing diverse kinds of entities. This tutorial will describe the language of probabilistic relational models (PRMs), a framework based on probabilistic graphical models, but extending them with the expressive power of object-relational languages. PRMs model the uncertainty over the attributes of objects in the domain as well as uncertainty over the existence of relations between objects. I will present techniques for automatically learning PRMs directly from a relational data set, and applications of these techniques to various tasks, such as: collective classification of an entire set of related entities; clustering a set of linked entities into coherent groups; and even predicting the existence of links between entities. I will describe applications of this framework to various tasks, including: recognizing regulatory and protein interactions in a cell from diverse types of genomic data; segmenting and recognizing objects in robot laser range scan data; and identifying the set of entities in a structured website and the relationships between them.
Presentation (PowerPoint File)
Video of Talk (RealPlayer File)
Copyright. All Rights Reserved.