Data Mining and Fusion
The massive amount of data available from different sources must be processed before it can be used to understand the world and how to act. We use both data fusion and data mining to process and interpret this data. Data fusion techniques provide a coherent picture of the world even when various data streams provide different types of data or conflicting data. For instance, we use a variety of probabilistic modeling techniques, including Bayesian models, probabilistic relational models, and probabilistic programming, to model many types of uncertainty about the various data streams and, ultimately, the world. Data mining provides automated techniques to identify and extract critical data elements that might otherwise be missed in massive datasets. We use both fully automated anomaly detection as well as mixed-initiative human-machine data exploration techniques as appropriate to rapidly explore many kinds of datasets for many different users.