Connections between Unsourced Multiple Access and Sparse Recovery
Multiple access communication has been studied extensively in information theory and wireless communications for several decades. Simultaneously, sparse recovery problems including compressed sensing, group testing, neighbor discovery in wireless networks, and data stream computing have also been studied in depth within their respective research communities. Although connections between multiple access and sparse recovery were pointed out as early as the 1980s, these fields have emerged mostly independently. A particular version of multiple access communication, called unsourced multiple access has become very popular recently due to its relevance for the Internet of Things. In this talk, we will show that there are strong connections between designing encoding and signal processing schemes for unsourced multiple access, and designing sensing matrices and recovery algorithms for sparse recovery problems in large dimensions. We will show how these connections can be gainfully exploited for designing algorithms with manageable complexity for both unsourced multiple access and sparse recovery. Our proposed techniques have applications in massive multiple access, neighbor discovery, lossy compressed sensing, heavy hitters problems, and group testing. This is joint work with Prof. Jean-Francois Chamberland and several former and current graduate students at Texas A&M University.
Krishna Narayanan is the Eric D. Rubin professor in the Electrical and Computer Engineering department at Texas A&M University. His recent research interests have been in massive multiple access for IoT, coded distributed computing, machine learning for joint source channel coding, and graph neural networks. He is a Fellow of IEEE and recently received the 2020 best paper award in data storage from IEEE communications society.