MURI : Value - centered Information Theory Year 2 IPR

This interdisciplinary project is developing a comprehensive set of principles for task-specific information extraction and information exploitation that can be used to design the next generation of autonomous and adaptive sensing systems. The significance of this research is that it addresses the widespread and longstanding problem of defining, assessing, and exploiting the value of information in active sensing systems. This year we report progress in twenty areas organized around three main thrusts: (1) learning and representation of high dimensional data, (2) distributed information fusion, and (3) active information exploitation. In the learning and representation thrust, progress ranges from assessing value of Kronecker representations of high-dimensional covariance matrices to learning to rank user preference data, an important task for human-in-the-loop decision systems. In the distributed information fusion thrust, progress is reported in assessing value of information in distributed information gathering and dimensionality reduction systems with application to sensor networks. In the active information exploitation thrust, progress is reported in information geometric trajectory planning, adversarial information collection, active learning in Bayes nets, and multistage adaptive estimation of sparse signals. Our future plans are to continue to develop linkages between these thrust areas, to further our development of fundamental theory for designing and evaluating distributed active information collection systems, and to account for human interactions in the sensing and processing loop.