Lecture by Krishna Narayanan, Texas A&M University

Title: Some recent results on coded distributed matrix multiplication

Speaker: Krishna Narayanan, Texas A&M University

Date; Monday, Aug. 19, 2019 4:15-5:15 pm,

Location: GJH, ECE

Abstract: We consider the problem of computing the product of two matrices in a distributed fashion using a coded matrix multiplication scheme with N worker nodes. Several classes of codes have been recently proposed to make the computation robust to stragglers and to nodes that return erroneous computations. In this talk, I will provide an overview of some of these coding schemes, and discuss their advantages and disadvantages – particularly, in terms of the numerical stability of the decoding algorithm.

I will then present our recent results on three closely related aspects of the problem – i) design of codes based on random Khatri-Rao products which have optimal recovery thresholds and a numerically stable decoding algorithm, ii) design of factored fountain codes which have near-optimal decoding thresholds with linear time decoding complexity, and iii) collaborative decoding which nearly doubles the decoding radius with high probability in the presence of random errors in the computation. I will mention some open problems.

Bio: Krishna R. Narayanan received the B.E. degree from Coimbatore Institute of Technology, Coimbatore, India, the M.S. degree from Iowa State University, Ames, IA, USA, and the Ph.D. degree from Georgia Institute of Technology, Atlanta, GA, USA, in 1992, 1994, and 1998, respectively. He is currently a Professor of Electrical and Computer Engineering with Texas A&M University, College Station, TX, USA. His research interests include coding theory, information theory, signal processing with applications to wireless communications, data storage, and data science. He served as an Area Editor for the coding theory and applications area of the IEEE Transactions on Communications from 2007 to 2011. He received the Distinguished Alumnus Award from Iowa State University and the 2007 Best Paper Award from the IEEE Signal Processing for Data Storage Society. He is an IEEE Fellow.

ALL ARE WELCOME