U.S. Department of Energy

Pacific Northwest National Laboratory

Learning Intelligent Assistants for Understanding Behavior

Thursday, August 27, 2015
Dr. Dragos Margineantu
In many cases data doesn’t come in cleanly labeled but can be interpreted as a result of actions of rational individuals. Can we build systems that sift through such data to assist experts in understanding what could be abnormal or anomalous, or in alerting for certain trends? We believe so! This talk will analyze how, by formulating adequately our research questions and by employing models that are capable of capturing domain interaction, learning can be made scalable via expert interaction. Next, we will discuss how our techniques can be employed for assisting experts in surveillance tasks such as intent recognition and detecting abnormal behavior. Finally, we will outline open research questions for usable expert-interactive learning.
Speaker Bio

Dragos Margineantu is a Senior Scientist and Technical Fellow at Boeing where his work and research interests are in machine learning, building systems that interact with humans and form teams with humans, and on applying machine learning techniques to practical tasks. In his role, he focuses on anomaly detection and rare event detection in multivariate sensor data; online learning; interactive learning and knowledge-based learning; object detection in images; machine learning tools for computer vision; intent analysis and intent recognition; inverse reinforcement learning, inverse planning; and active learning and ensemble learning (applied and integrated). Margineantu has chaired numerous conferences and workshops, most recently KDD-2015 Industry and Government Track. Prior to Boeing, he served as Director, Machine Learning at First Data Corporation. He received his Ph.D. in Computer Science / Machine Learning from Oregon State University in 2001.

| Pacific Northwest National Laboratory