Dragos D. Margineantu

Current Research

My work focuses on both: the research and engineering of intelligent systems.

In research, my interests are in machine learning and probabilistic programming. Lately, I am mostly focused on addressing the research questions that lead to building robustness into learned models and intelligent systems. Estimating the risk (and bounding the risk) of learned models' output at time of deployment, learning calibrated probabilities, and probabilistic inference & learning for multifaceted learning (learning "to decide in more than one way") are the major research questions that my group and I try to address around robustness.
I am also interested in the fundamental research questions that we need to address to build trustable intelligent systems that are required to interact with human experts to make decisions. Research questions on inverse reinforcement learning for anomalous action detection, and novel methods for anomaly detection and change detection are at the center of our research.

In the engineering of intelligent systems I am particularly interested in architectures of decision systems that are driven by scalability and can support learning, inference and knowledge systems implementations.
I am leading project teams that develop autonomous airplanes, systems health monitoring, surveillance and anomaly detection systems. That's why I am particularly interested in embedded AI systems and the details of the how the functions of intelligent systems are decomposed for the impolementation and deployment of practical solutions.
The representations of domain and commonsense knowledge that support learning and inference are crucial for practical AI systems. That's why our group has excelled in analyzing, designing, and implementing such representations for autonomous systems, for the design of complex systems and for modern anomaly detection systems.

Most predictions should ultimately lead to decisions and actions and those decisions and actions require have costs and risks associated with them. Therefore I am very interested in learning and decision making techniques that deal with costs, budgets and risks (typically these are non-uniform functions). Cost-sensitive learning, active learning, and hierarchical learning are typically required in any practical application of machine learning algorithms.

I am also interested and working on statistical tests and validation methods for decision systems, learned models, and learning algorithms.

General categories of machine learning methods that myself and my colleagues have implemented and have experience with: ensemble learning, active learning, semi-supervised learning, clustering, deep learning, inverse reinforcement learning, sequential decision making and reinforcement learning.

In general, I am interested in listening and discussing novel research ideas in machine learning, probabilistic inference, inverse reinforcement learning, and game theory.