Title:  Dynamic Low-rank Matrix Recovery: Theory and Applications

Committee: 

Dr. Davenport, Advisor  

Dr. Rozell, Chair

Dr. Romberg

Abstract:

The objective of the proposed research is to provide theoretical understanding of dynamic low-rank matrix recovery. Although the benefits of harnessing dynamics in low-rank matrix recovery has been observed in various applications, theoretical understanding and justification for exploiting dynamics is limited. In this proposal we mainly aim to summarize some prototypical dynamic models in the context of low-rank matrix recovery and study their theoretical properties, including recovery guarantees and algorithmic convergence. We aim to answer questions such as: What kind of reduction in the sample complexity is possible by exploiting dynamic structure in the underlying matrix and what are the recovery error guarantees compared to the corresponding guarantees for the static baseline cases? We also aim to provide numerical simulations to validate our analysis and real-world experiments to show their empirical effectiveness.