Research

My interests lie broadly in signal processing, machine learning, optimization, and statistics, with a focus on the two areas below:

Adaptive sensing, learning, and planning

I am developing methods that continuously incorporate information learned during processing, feeding it back to shape future actions. Most conventional methods, in contrast, are designed around prior knowledge with little capacity for change. I have shown that for sparse signals, e.g. targets in a surveillance context, adaptive sensing over multiple stages can result in much more focused use of sensing resources (with analytical guarantees) compared to the non-adaptive alternative (as in compressive sensing). This work leads to extensions to time-varying signals and applications in spectrum sensing, scientific data collection, and business management.

Optimization in signal processing and machine learning

The improvement in algorithms and processing power has made available more sophisticated uses of optimization. I am exploring new methods for exploiting low-dimensional structure in high-dimensional data, for example to learn correlations between nodes (e.g. people) in a graphical model (e.g. a social network) in a distributed manner, limiting computation and communication to local neighborhoods in the graph. I have also applied optimization to the design of sparse digital filters, variable selection in high-dimensional regression, and robust synthetic aperture radar imaging.