ECE Seminar - Towards Anarchic Federated Learning
Department of Electrical and Computer Engineering
The University of Alabama, Tuscaloosa, AL
Date: Wednesday, March 27, 2024
Time: 2:00PM - 3:00PM
Location: North Lawn Hall 1018
Speaker: Dr. Xiaowen Gong
Associate Professor, Department of Electrical and Computer Engineering, Auburn University
The University of Alabama, Tuscaloosa, AL
Date: Wednesday, March 27, 2024
Time: 2:00PM - 3:00PM
Location: North Lawn Hall 1018
Speaker: Dr. Xiaowen Gong
Associate Professor, Department of Electrical and Computer Engineering, Auburn University
Title: Towards Anarchic Federated Learning
Abstract: As an emerging topic that has received tremendous research in the past few years, federated learning (FL) has numerous promising applications in networked intelligent systems, such as connected and autonomous vehicles, collaborative robots. Existing work on FL often assume that clients participate in the learning process with some particular pattern (such as balanced participation), and/or in a synchronous manner, and/or with the same number of local iterations, while these assumptions can be hard to hold in practice.
In this talk, we will present our recent research on anarchic FL (AFL), which gives maximum freedom to clients, so that clients can participate in FL flexibly and efficiently according to their heterogeneous and time-varying computation and communication capabilities. In particular, we will discuss three studies in this research direction: 1) anarchic FL with delayed gradient averaging where clients can compute gradients and communicate gradients in parallel; 2) anarchic bilevel FL where clients perform bilevel learning; 3) anarchic decentralized FL where clients communicate in a decentralized structure. We will talk about various challenges in the algorithm design and convergence analysis of the AFL algorithms in these studies, and the corresponding techniques that address these challenges. We will also discuss future research directions.
Bio: Xiaowen Gong is currently a Godbold Associate Professor in the Department of Electrical and Computer Engineering (ECE) at Auburn University. He received his PhD degree in Electrical Engineering from Arizona State University (ASU) in 2015. From 2015 to 2016, he was a postdoctoral researcher in the Department of ECE at The Ohio State University. His research interests are in the areas of wireless networks and their applications, with current focuses on machine learning and AI in wireless networks. He received IEEE Internet of Things Journal Best Paper Runner-up Award in 2022 as a co-author, IEEE INFOCOM 2014 Runner-up Best Paper Award as a co-author, ASU ECEE Palais Outstanding Doctoral Student Award in 2015, and NSF CAREER Award in 2022. He has served as an Associate Editor for IEEE Transactions on Wireless Communications, a Guest Editor for IEEE Transactions on Network Science and Engineering, and a Lead Guest Editor for IEEE Open Journal of the Communications Society.
Abstract: As an emerging topic that has received tremendous research in the past few years, federated learning (FL) has numerous promising applications in networked intelligent systems, such as connected and autonomous vehicles, collaborative robots. Existing work on FL often assume that clients participate in the learning process with some particular pattern (such as balanced participation), and/or in a synchronous manner, and/or with the same number of local iterations, while these assumptions can be hard to hold in practice.
In this talk, we will present our recent research on anarchic FL (AFL), which gives maximum freedom to clients, so that clients can participate in FL flexibly and efficiently according to their heterogeneous and time-varying computation and communication capabilities. In particular, we will discuss three studies in this research direction: 1) anarchic FL with delayed gradient averaging where clients can compute gradients and communicate gradients in parallel; 2) anarchic bilevel FL where clients perform bilevel learning; 3) anarchic decentralized FL where clients communicate in a decentralized structure. We will talk about various challenges in the algorithm design and convergence analysis of the AFL algorithms in these studies, and the corresponding techniques that address these challenges. We will also discuss future research directions.
Bio: Xiaowen Gong is currently a Godbold Associate Professor in the Department of Electrical and Computer Engineering (ECE) at Auburn University. He received his PhD degree in Electrical Engineering from Arizona State University (ASU) in 2015. From 2015 to 2016, he was a postdoctoral researcher in the Department of ECE at The Ohio State University. His research interests are in the areas of wireless networks and their applications, with current focuses on machine learning and AI in wireless networks. He received IEEE Internet of Things Journal Best Paper Runner-up Award in 2022 as a co-author, IEEE INFOCOM 2014 Runner-up Best Paper Award as a co-author, ASU ECEE Palais Outstanding Doctoral Student Award in 2015, and NSF CAREER Award in 2022. He has served as an Associate Editor for IEEE Transactions on Wireless Communications, a Guest Editor for IEEE Transactions on Network Science and Engineering, and a Lead Guest Editor for IEEE Open Journal of the Communications Society.
ECE Seminar - Empowering Autonomous Vehicles to Perceive the Invisible
Department of Electrical and Computer Engineering
The University of Alabama, Tuscaloosa, AL
Date: Wednesday, Jan. 31, 2024
Time: 2:00PM - 3:00PM
Location: North Lawn Hall 1018
Speaker: Dr. Xinyu Zhang
Associate Professor, Department of Electrical and Computer Engineering, University of California, San Diego
The University of Alabama, Tuscaloosa, AL
Date: Wednesday, Jan. 31, 2024
Time: 2:00PM - 3:00PM
Location: North Lawn Hall 1018
Speaker: Dr. Xinyu Zhang
Associate Professor, Department of Electrical and Computer Engineering, University of California, San Diego
Title: Empowering Autonomous Vehicles to Perceive the Invisible
Abstract: Self-driving vehicles are poised to overhaul transportation networks in the imminent future. To handle complex navigational tasks, these vehicles increasingly rely on sophisticated sensor systems--cameras, lidar, and radar--to interpret the environment. Despite the impressive strides made in computer vision, optical sensors like cameras and lidar often struggle when confronted with poor visibility or harsh weather conditions. Radio frequency (RF) sensors, especially radar, are less susceptible to these disruptions but traditionally suffer from significantly lower resolution. In this talk, I will introduce radical sensing paradigms and smart infrastructures designed to overcome these inherent limitations. I will first unveil how passive and intelligent metasurfaces, integrated into roadside infrastructure, can convey vital data to automotive radars, while granting vehicles the ability to 'see around corners.' I will then discuss how to craft high-definition perception from sparse radar signals using phased array synthesis and cutting-edge generative models. By enhancing both the infrastructure and radar sensing models, we can empower vehicles to perceive the previously invisible with unprecedented coverage and resolution.
Bio: Xinyu Zhang is an Associate Professor in the Department of Electrical and Computer Engineering at the University of California San Diego. He received his Ph.D. in Computer Science and Engineering from the University of Michigan in 2012. His research interest lies in wireless networking and ubiquitous sensing, and more specifically in (i) next-generation wireless network architectures and protocols; (ii) ubiquitous systems that leverage wireless signals to sense micro-locations and micro-activities with near-vision precision. He is the recipient of two ACM MobiCom Best Paper Awards (2011 and 2020), Communications of the ACM Research Highlight (2018), ACM SIGMOBILE Research Highlight (2018), NSF CAREER Award (2014), Google Research Award (2017, 2018, 2020), Sony Research Award (2018, 2020), and ACM SIGMOBILE RockStar Award in 2023. He served as the TPC chair for ACM MobiCom 2019, IEEE SECON 2017, co-chair of the NSF millimeter-wave research coordination network, and Associate Editor for IEEE Transactions on Mobile Computing from 2017 to 2020.
Abstract: Self-driving vehicles are poised to overhaul transportation networks in the imminent future. To handle complex navigational tasks, these vehicles increasingly rely on sophisticated sensor systems--cameras, lidar, and radar--to interpret the environment. Despite the impressive strides made in computer vision, optical sensors like cameras and lidar often struggle when confronted with poor visibility or harsh weather conditions. Radio frequency (RF) sensors, especially radar, are less susceptible to these disruptions but traditionally suffer from significantly lower resolution. In this talk, I will introduce radical sensing paradigms and smart infrastructures designed to overcome these inherent limitations. I will first unveil how passive and intelligent metasurfaces, integrated into roadside infrastructure, can convey vital data to automotive radars, while granting vehicles the ability to 'see around corners.' I will then discuss how to craft high-definition perception from sparse radar signals using phased array synthesis and cutting-edge generative models. By enhancing both the infrastructure and radar sensing models, we can empower vehicles to perceive the previously invisible with unprecedented coverage and resolution.
Bio: Xinyu Zhang is an Associate Professor in the Department of Electrical and Computer Engineering at the University of California San Diego. He received his Ph.D. in Computer Science and Engineering from the University of Michigan in 2012. His research interest lies in wireless networking and ubiquitous sensing, and more specifically in (i) next-generation wireless network architectures and protocols; (ii) ubiquitous systems that leverage wireless signals to sense micro-locations and micro-activities with near-vision precision. He is the recipient of two ACM MobiCom Best Paper Awards (2011 and 2020), Communications of the ACM Research Highlight (2018), ACM SIGMOBILE Research Highlight (2018), NSF CAREER Award (2014), Google Research Award (2017, 2018, 2020), Sony Research Award (2018, 2020), and ACM SIGMOBILE RockStar Award in 2023. He served as the TPC chair for ACM MobiCom 2019, IEEE SECON 2017, co-chair of the NSF millimeter-wave research coordination network, and Associate Editor for IEEE Transactions on Mobile Computing from 2017 to 2020.
ECE Signal Processing Seminar/IEEE AESS Distinguished Lecture - A Dual-Blind Deconvolution Perspective of Integrated Sensing and Communications
Department of Electrical and Computer Engineering
The University of Alabama, Tuscaloosa, AL
Date: Tuesday, July 11, 2023
Time: 1:30PM - 2:30PM
Location: NERC 1012
Speaker: Dr. Kumar Vijay Mishra
Senior Fellow at the United States Army Research Laboratory (ARL), Adelphi, MD
The University of Alabama, Tuscaloosa, AL
Date: Tuesday, July 11, 2023
Time: 1:30PM - 2:30PM
Location: NERC 1012
Speaker: Dr. Kumar Vijay Mishra
Senior Fellow at the United States Army Research Laboratory (ARL), Adelphi, MD
Title: A Dual-Blind Deconvolution Perspective of Integrated Sensing and Communications
Abstract: Recent interest in integrated sensing and communications (ISAC) has led to the design of novel signal processing techniques to recover information from an overlaid radar-communications signal. In this talk, we present a general spectral coexistence scenario, wherein the channels and transmit signals of both radar and communications systems are unknown at the receiver. In this dual-blind deconvolution (DBD) problem, a common receiver admits a multi-carrier wireless communications signal that is overlaid with the radar signal reflected off multiple targets. The communications and radar channels are represented by continuous-valued range-time and Doppler velocities of multiple transmission paths and multiple targets. We exploit the sparsity of both channels to solve the highly ill-posed DBD problem by casting it into a sum of multivariate atomic norms (SoMAN) minimization. We devise a semidefinite program to estimate the unknown target and communications parameters using the theories of positive-hyperoctant trigonometric polynomials (PhTP). Our theoretical analyses show that the minimum number of samples required for perfect recovery scale logarithmically with the maximum of the radar targets and communications paths rather than their sum. We show that our SoMAN method and PhTP formulations are also applicable to more general scenarios such as unsynchronized transmission, the presence of noise, and multiple emitters. We also examine this problem using extremal functions from the Beurling-Selberg interpolation theory. Toward the end of the talk, we will briefly touch upon our research on other ISAC topics.
Bio: Kumar Vijay Mishra (S’08-M’15-SM’18) obtained a Ph.D. in electrical engineering and M.S. in mathematics from The University of Iowa in 2015, and M.S. in electrical engineering from Colorado State University in 2012, while working on NASA’s Global Precipitation Mission Ground Validation (GPM-GV) weather radars. He received his B. Tech. summa cum laude (Gold Medal, Honors) in electronics and communication engineering from the National Institute of Technology, Hamirpur (NITH), India in 2003. He is currently Senior Fellow at the United States Army Research Laboratory (ARL), Adelphi; Technical Adviser to Singapore-based automotive radar start-up Hertzwell and Boston-based imaging radar startup Aura Intelligent Systems; and honorary Research Fellow at SnT - Interdisciplinary Centre for Security, Reliability and Trust, University of Luxembourg. Previously, he had research appointments at Electronics and Radar Development Establishment (LRDE), Defence Research and Development Organisation (DRDO) Bengaluru; IIHR - Hydroscience & Engineering, Iowa City, IA; Mitsubishi Electric Research Labs, Cambridge, MA; Qualcomm, San Jose; and Technion - Israel Institute of Technology.
Dr. Mishra is the Distinguished Lecturer of the IEEE Communications Society (2023-2024), IEEE Aerospace and Electronic Systems Society (AESS) (2023-2024), and IEEE Future Networks Initiative (2022). He is the recipient of the IET Premium Best Paper Prize (2021), U. S. National Academies Harry Diamond Distinguished Fellowship (2018-2021), American Geophysical Union Editors' Citation for Excellence (2019), Royal Meteorological Society Quarterly Journal Editor's Prize (2017), Viterbi Postdoctoral Fellowship (2015, 2016), Lady Davis Postdoctoral Fellowship (2017), DRDO LRDE Scientist of the Year Award (2006), NITH Director’s Gold Medal (2003), and NITH Best Student Award (2003). He has received Best Paper Awards at IEEE MLSP 2019 and IEEE ACES Symposium 2019.
Dr. Mishra is Chair (2023-present) of the Synthetic Apertures Technical Working Group of the IEEE Signal Processing Society (SPS) and Vice-Chair (2021-present) of the IEEE Synthetic Aperture Standards Committee, which is the first SPS standards committee. He is the Vice Chair (2021-2023) and Chair-designate (2023-2026) of the International Union of Radio Science (URSI) Commission C. He has been an elected member of three technical committees of IEEE SPS: SPCOM, SAM, and ASPS, and IEEE AESS Radar Systems Panel. Since 2020, he has been Associate Editor of IEEE Transactions on Aerospace and Electronic Systems, where he was awarded Outstanding Editor recognition in 2021. He has been a lead/guest editor of several special issues in journals such as IEEE Signal Processing Magazine, IEEE Journal of Selected Topics in Signal Processing, and IEEE Journal on Selected Areas in Communications. He is the lead co-editor of three upcoming books on radar: Signal Processing for Joint Radar-Communications (Wiley-IEEE Press), Next-Generation Cognitive Radar Systems (IET Press Radar, Electromagnetics & Signal Processing Technologies Series), and Advances in Weather Radar Volumes 1, 2 & 3 (IET Press Radar, Electromagnetics & Signal Processing Technologies Series). His research interests include radar systems, signal processing, remote sensing, and electromagnetics.