Prof. Nikos Sidiropoulos (University of Minnesota)
Title: “Tensor Decomposition Theory and Algorithms in the Era of Big Data”
Host: Jakub Szefer
Abstract: Tensor decomposition is a fascinating cross-disciplinary area that builds upon linear and multilinear algebra, complexity, optimization, and diverse applications – from chemistry to data mining and machine learning. Tensors (data indexed by three or more indices) have important similarities but also striking differences with matrices (indexed by row and column). We will begin with a retrospective on tensors with emphasis on low-rank decomposition / approximation and its properties, including uniqueness, algorithms, complexity, and performance bounds. Most of these results and tools are geared towards small- to moderate-size tensors that can comfortably fit in main memory. In modern applications, such as brain imaging, web and social data mining, and recommender systems, this is no longer true. The tensors are very large, typically very sparse, and may be distributed in cloud storage. These characteristics bring up new questions and challenges, such as whether it is possible to work with low-dimensional sketches of the data without losing identifiability, or how to parallelize big tensor computations in a memory, storage, computation and communication-friendly way. We will talk about multilinear compressed sensing, its oracle properties, and how it can be used for Hadoop-type parallel computation; and efficient parallel algorithms using the alternating direction method of multipliers that are well-suited for shared memory and multi-core architectures.
Bio: Nikos Sidiropoulos received the Diploma in Electrical Engineering from the Aristotelian University of Thessaloniki, Greece, and M.S. and Ph.D. degrees in Electrical Engineering from the University of Maryland–College Park, in 1988, 1990 and 1992, respectively. He served as assistant professor at the University of Virginia, associate professor at the University of Minnesota, and professor at TU Crete, Greece. Since 2011, he has been at the University of Minnesota, where he currently holds an ADC Chair in digital technology. His research spans topics in signal processing theory and algorithms, optimization, communications, and factor analysis - with a long-term interest in tensor decomposition and its applications. His current focus is primarily on signal and tensor analytics for learning from big data. He received the NSF/CAREER award in 1998, and the IEEE Signal Processing (SP) Society Best Paper Award in 2001, 2007, and 2011. He served as IEEE SP Society Distinguished Lecturer (2008-2009), and as Chair of the IEEE Signal Processing for Communications and Networking Technical Committee (2007-2008). He received the 2010 IEEE SP Society Meritorious Service Award, and the 2013 Distinguished Alumni Award from the Dept. of ECE, University of Maryland. He is a Fellow of IEEE (2009) and a Fellow of EURASIP (2014).