Asst Professor, ECE
High Dimensional Statistical Signal Processing and Data Analysis, Energy Efficient Sketching and Sampling For Statistical Inference, Compressive Sensing and Sparse Estimation, Tensor Methods, Convex and Non-Convex Optimization, Optical Signal Processing and High Resolution Imaging, Statistical Learning.
The central goal of Prof. Pal’s research is to design and analyze new energy-efficient sensing paradigms coupled with computationally efficient robust algorithms for information processing of high dimensional signals, and understand their fundamental performance limits. She exploits the laws of physics associated with signal generation, acquisition and propagation, along with available domain knowledge (often characterized in terms of statistical priors) to judiciously design new structured sampling techniques and robust estimation/detection algorithms that can work together to estimate parameters of interest from highly undersampled, noisy and corrupted data. In her past and ongoing work in signal processing for sensor arrays, she developed new sampling geometries, namely nested and coprime samplers, that consume significantly lower power and require fewer sensors than existing methods, overcoming long-standing performance bottlenecks. In the context of compressive sensing and sparse estimation, she has developed a new “correlation-aware” technique for sparse estimation which explicitly utilizes spatio-temporal correlation present in multichannel data. Using this new framework, she has shown that it is fundamentally possible to increase (by an order of magnitude) the level of recoverable sparsity beyond what is achievable by state-of-the-art compressive sensing algorithms that typically fail to exploit such correlation priors.
More recently, she is working on understanding the role of samplers in achieving fundamental limits of covariance compression and covariance-driven statistical inference; developing new algorithms for solving highly underdetermined inverse problems in high resolution imaging that reveal fundamental interplays between sample size, resolution, and priors obtained from statistical and physical signal models; and designing and analyzing sampling techniques for tensor methods in statistical machine learning.
Piya Pal received her Ph.D. in Electrical Engineering from California Institute of Technology in 2013. Prior to her appointment at UC San Diego, she was an Assistant Professor of Electrical and Computer Engineering at the University of Maryland, College Park where she was also affiliated with the Institute for Systems Research. Her doctoral thesis titled “New directions in sparse sampling and estimation for underdetermined systems” was awarded the 2014 Charles and Ellen Wilts Prize for Outstanding Thesis in Electrical Engineering at Caltech. She received an NSF CAREER Award in 2016 to pursue her research in “Smart Sampling and Correlation-Driven Inference for High Dimensional Signals."