Title:  Learning to Adapt under Practical Sensing Constraints

Committee: 

Dr. Davenport, Advisor

Dr. Rozell, Chair

Dr. Romberg

Abstract:

The objective of the proposed research is to combine theory in signal processing and machine learning to develop techniques for adaptive signal acquisition which can cope with practical measurement constraints. By leveraging structured data models such as sparsity, intelligent sampling schemes can enable higher quality estimation with less labeled data in diverse applications such as imaging, recommendation systems, information retrieval, and psychometric studies. The proposed work will target three aims: (i) investigate the best selection of observations for linear regression, (ii) develop theory for localizing a point via sequentially-chosen paired comparisons, and (iii) design methods for adaptive measurement selection in one-bit constrained sensing.