Arizona State University Network Science Seminar Series

Upcoming Seminar: On addressing uncertainty and high-dimensionality in optimization and variational inequality problems: self-tuned stepsizes, and randomized block coordinate schemes

Speaker Farzad Yousefian (Oklahoma State University)
Date 1:30 p.m., Mar 17, 2017
Location GWC 487
Short Bio
Farzad Yousefian is currently an assistant professor in the school of Industrial Engineering and Management at Oklahoma State University. Before joining OSU, he was a postdoctoral researcher in the Department of Industrial and Manufacturing Engineering at Penn State. He obtained his Ph.D. in industrial engineering from the University of Illinois at Urbana-Champaign in 2013. His thesis is focused on the design, analysis, and implementation of stochastic approximation methods for solving optimization and variational problems in nonsmooth and uncertain regimes. His current research interests lie in the development of efficient algorithms to address ill-posed stochastic optimization and equilibrium problems arising from machine learning and multi-agent systems. He is the recipient of the best theoretical paper award in the 2013 Winter Simulation Conference.
A wide range of emerging applications in machine learning, signal processing, and multi-agent systems result in optimization, and more generally variational inequality problems. Such models are often complicated by uncertainty, and/or high-dimensionality. In the first part of this talk, we consider stochastic mirror descent methods for solving stochastic convex optimization problems. It has been discussed that the performance of this class of methods is very sensitive to the choice of the stepsize sequence. Motivated by this gap, we present a unifying self-tuned update rule for the stepsize sequence such that: (i) it is characterized in terms of problem parameters and algorithm’s settings; and (ii) under this update rule, a suitably defined error metric is minimized. We present the performance of this update rule for the soft margin linear SVM problem over different large data sets.

In the second part of the talk, motivated by multi-user optimization problems and non-cooperative Nash games in uncertain regimes, we consider stochastic Cartesian variational inequalities where the number of the component sets is huge. We develop a randomized block stochastic mirror-prox (B-SMP) algorithm, where at each iteration only a randomly selected block coordinate of the solution is updated through implementing two consecutive projection steps. The convergence analysis of the B-SMP method equipped with rate statements will be presented.


Title Speaker Time Location
Subspace Detection with Applications Louis Scharf (Colorado State University) 1:30 p.m., Jan 26th, 2017 GWC 487
An Informational Perspective on Uncertainty in Control Gireeja Ranade (Microsoft Research, Redmond) 1:30 p.m., Feb 27th, 2017 GWC 409
Orthogonal precoding for sidelobe suppression in DFT-based systems using block reflectors Vaughan Clarkson (University of Queensland) 1:30 p.m., Mar 3rd, 2017 GWC 487
On addressing uncertainty and high-dimensionality in optimization and variational inequality problems: self-tuned stepsizes, and randomized block coordinate schemes Farzad Yousefian (Oklahoma State University)) 1:30 p.m., Mar 17th, 2017 GWC 487

Subspace Detection with Applications

Speaker Louis Scharf (Colorado State University)
Date 1:30 p.m., Jan 26, 2017
Location GWC 487
Short Bio
Louis Scharf received his Ph.D. from the University of Washington, Seattle. From 1971 to 1982 he served as Professor of Electrical Engineering and Statistics at CSU. From 1982 to 1985 he was Professor and Chairman of Electrical and Computer Engineering at the University of Rhode Island, Kingston. From 1985 to 2000 he was Professor of Electrical and Computer Engineering at the University of Colorado, Boulder. In January 2001, Professor Scharf rejoins Colorado State University as Professor of Electrical and Computer Engineering, and Statistics.

Professor Scharf has held several visiting positions here and abroad. He has developed particularly close ties with Ecole Superieure d'Electricite (Gif-sur-Yvette), Ecole Nationale Superieure des Telecommunications (Paris), and EURECOM (Nice). He is a recognized expert in statistical signal processing, as it applies to adaptive radar, sonar, and wireless communication. His most important contributions to date are to invariance theories for detection and estimation; matched and adaptive subspace detectors for radar, sonar, and data communication; and canonical decompositions for reduced dimensional filtering and quantizing. His current interests are in rapidly-adaptive receiver design for space-time signal processing in the wireless communication channel.

Professor Scharf is a Fellow of IEEE. He chairs the Fellow Committee for the IEEE Signal Processing Society, and serves on its Technical Committees for Theory and Methods and for Sensor Arrays and Multichannel Signal Processing. He has received numerous awards for his research contributions to statistical signal processing, including an IEEE Distinguished Lectureship, an IEEE Third Millenium Medal, and the Technical Achievement Award from the IEEE Signal Processing Society.
(To be Updated)

An informational perspective on uncertainty in control

Speaker Gireeja Ranade (Microsoft Research, Redmond)
Date 1:30 p.m., Feb 27, 2017
Location GWC 409
Short Bio
Gireeja Ranade is a postdoctoral researcher at Microsoft Research, Redmond. Before this she was a lecturer in EECS at UC Berkeley working on designing and teaching the pilot version of novel lower-division EECS classes (16AB). She received an MS and PhD in EECS from UC Berkeley and an SB in EECS from MIT. She has worked on topics in brain-machine interfaces, information theory, control theory, wireless communications and crowdsourcing.
Developing high-performance cyber-physical systems requires a deep understanding of how uncertainty and unpredictability impair performance. In this talk, I discuss some theoretical perspectives to understand uncertainty in systems as well as practical protocols to mitigate it. I will first introduce a notion of "control capacity," which parallels the notion of Shannon communication capacity, and provides a fundamental limit on the ability to stabilize a system with random time-varying parameters (modeled as multiplicative noise). Further, it can be used to quantify the value of side-information in control. We contrast systems with noisy actuation (e.g., when motors on a drone cannot precisely execute control actions) to noisy sensing (e.g., miscalibrated cameras). In the first case, we show that linear control strategies are optimal, while in the second, we show that non-linear strategies can outperform them. Further, we use techniques from information-theory and probability-theory to bound the improvement that non-linear strategies can bring. Finally, I will shift from quantifying the effect of uncertainty to methods for reducing uncertainty. With the aim of enabling industrial automation, I will discuss the development of highly-reliable low-latency wireless communication protocols for machine-to-machine communication. The talk will include joint work with Jian Ding, Yuval Peres, Govind Ramnarayan, Anant Sahai, Sahaana Suri, Vasuki Narasimha Swamy, and Alex Zhai.

Orthogonal precoding for sidelobe suppression in DFT-based systems using block reflectors

Speaker Vaughan Clarkson (University of Queensland)
Date 1:30 p.m., Mar 3, 2017
Location GWC 487
Short Bio
Bachelor of Science (Mathematics), UQ, 1989
Bachelor of Engineering (Computer Systems; Hons I), UQ, 1990
Doctor of Philosophy, ANU, 1997
Sidelobe suppression has always been an important part of crafting communications signals to keep interference with users of adjacent spectrum to a minimum. Systems based on the discrete Fourier transform, such as orthogonal frequency-division multiplexing (OFDM) and single-carrier frequency-division multiple access (SC-FDMA) are especially prone to out-of-band power leakage. Although many techniques have been proposed to suppress sidelobes in DFT-based systems, a satisfactory balance between computational complexity and out-of-band power leakage has remained elusive.

Orthogonal precoding is a promising, linear technique in which the nullspace of a precoding matrix with orthonormal columns is designed to suppress the sidelobes. Orthogonal precoders have been proposed that yield excellent out-of-band suppression. However, they suffer from high arithmetic complexity—quadratic in the number of active subcarriers—which has limited their application.

In this talk, we find that the arithmetic complexity can be made linear instead of quadratic if a block reflector is used to perform the precoding instead of an otherwise unstructured unitary transformation. There is no penalty to be paid in achieved bit-error rate. We show by numerical simulation that the penalty in peak-to-average power ratio is also very small for OFDM.