Dr. Faradonbeh's research focuses on different aspects of algorithms that can bridge from limited data to reliable outcomes, yet enjoying computational friendliness. That includes three major categories focusing on design and analysis of algorithms, with applications in precision medicine, healthcare management, online advertisement, intelligent tutoring, precision agriculture, and system engineering.
The first is the study of methods that can accurately learn many parameters from structured data, as well as experimentation for collecting useful data. Commonly, different technical tools from high-dimensional statistics and Bayesian machine learning are utilized to address issues such as spatiotemporal dependence and compound uncertainties.
The second category is that of optimal decision-making in settings such as network systems and patient scheduling. The important issues include NP-hardness and guaranteed reliability, and approaches from combinatorial optimization and stochastic control are used.
Projects in the third category consist of reinforcement learning algorithms for real-time data-driven prescriptions, e.g., for tutor recommendation or dynamic treatments. Along these axes, fundamentally applicable settings are adopted together with a variety of frameworks from online learning and stochastic analysis, among others.