Abstract : 6G presents an opportunity to reflect on the fundamentals of wireless communication, as it becomes more and more difficult to estimate channels in high-mobility/high-Doppler environments when information signaling and signal processing are carried out in the traditional time-frequency (TF) domain. Also, the convergence of communication and radar sensing in 6G and beyond (inspired by the developments in intelligent transportation systems) has focused research attention on the design of waveforms that can simultaneously support both communication and radar sensing. This talk will focus on new research on communication and radar sensing using pulsones, a promising family of waveforms for this purpose. Pulsones are time domain realizations of quasi-periodic pulses in the delay-Doppler (DD) domain, parameterized by the delay period of the waveform with Zak theory providing the formal mathematical framework. Zak theory is to linear time-varying (LTV) systems as Fourier theory is to linear time-invariant (LTI) systems. As a communication waveform, pulsones offer the beneficial attributes of non-fading and predictability of input-output relation, leading to their robust bit error performance in rapidly time-varying channels. They are also natural waveforms for radar sensing as they provide good localization characteristics in the delay-Doppler domain. The optimum operating regime for both communication and radar sensing turns out to be the same, i.e., when the delay period and Doppler period of the waveform are larger than the delay spread and Doppler spread of the channel, respectively, a condition referred to as the crystallization condition.
Bio : A. Chockalingam received the B.E. (Honors) degree in ECE from P.S.G. College of Technology, Coimbatore in 1984 and the M. Tech degree in E & ECE from IIT, Kharagpur in 1985. In 1993, he obtained the Ph.D. degree in ECE from IISc, Bangalore. From Dec.1993 to May 1996, he was a Postdoctoral Fellow and an Assistant Project Scientist with the Department of ECE, UC San Diego. From May 1996 to Dec. 1998, he was with Qualcomm, San Diego, as a Staff Engineer/Manager. He joined the faculty of IISc in Dec. 1998 as an Assistant Professor in the Department of ECE, where he is currently a Professor working in the area of wireless communications.
Abstract : Many applications from camera arrays to sensor networks require efficient compression and processing of correlated data, which in general is collected in a distributed fashion. While information-theoretic foundations of distributed compression are well investigated, the impact of theory in practice has been somewhat limited. As the field of data compression is undergoing a transformation with the emergence of learning-based techniques, machine learning is becoming an important tool to reap the long-promised benefits of distributed compression. In this talk, we review the recent progress in the broad area of learned distributed compression, focusing on images as well as abstract sources. In particular, we discuss approaches that provide interpretable results operating close to information-theoretic bounds. We also discuss how learned distributed compression can impact multi-hop communications.
Bio : Elza Erkip is an Institute Professor in the Electrical and Computer Engineering Department at New York University Tandon School of Engineering. She received the B.S. degree in Electrical and Electronics Engineering from Middle East Technical University, Ankara, Turkey, and the M.S. and Ph.D. degrees in Electrical Engineering from Stanford University, Stanford, CA, USA. Her research interests are in information theory, communication theory, and wireless communications.
Dr. Erkip is a member of the Science Academy of Turkey and is a Fellow of the IEEE. She received the NSF CAREER award in 2001, the IEEE Communications Society WICE Outstanding Achievement Award in 2016, the IEEE Communications Society Communication Theory Technical Committee (CTTC) Technical Achievement Award in 2018, and the IEEE Communications Society Edwin Howard Armstrong Achievement Award in 2021. She was the Padovani Lecturer of the IEEE Information Theory Society in 2022. Her paper awards include the IEEE Communications Society Stephen O. Rice Paper Prize in 2004, the IEEE Communications Society Award for Advances in Communication in 2013 and the IEEE Communications Society Best Tutorial Paper Award in 2019. She was a member of the Board of Governors of the IEEE Information Theory Society 2012-2020, where she was the President in 2018. She was a Distinguished Lecturer of the IEEE Information Theory Society from 2013 to 2014.
Abstract : Distributed inference refers to the synergistic combination of information gathered by various knowledge sources and sensors to provide inference regarding a phenomenon of interest. This fascinating field has evolved over the past four decades and is being applied to a wide variety of fields such as military command and control, robotics, image processing, air traffic control, medical diagnostics, pattern recognition, environmental monitoring, IoT and smart cities. This talk will present an overview of the field, present some recent research results, illustrate its utility by means of some examples and conclude with some open problems.
Bio : Pramod K. Varshney was born in Allahabad, India, in 1952. He received the B.S. degree in electrical engineering and computer science (with highest honors), and the M.S. and Ph.D. degrees in electrical engineering from the University of Illinois at Urbana-Champaign in 1972, 1974, and 1976 respectively. Since 1976 he has been with Syracuse University, Syracuse, NY where he is currently a Distinguished Professor of Electrical Engineering and Computer Science. His current research interests are in distributed sensor networks and data fusion, detection and estimation theory, wireless communications, machine learning, AI and radar.
Dr. Varshney was the recipient of the 1981 ASEE Dow Outstanding Young Faculty Award. He was elected to the grade of Fellow of the IEEE in 1997 for his contributions in the area of distributed detection and data fusion. In 2000, he received the Third Millennium Medal from the IEEE and Chancellor’s Citation for exceptional academic achievement at Syracuse University. He is the recipient of the IEEE 2012 Judith A. Resnik Award. He received an honorary Doctor of Engineering degree from Drexel University in 2014, ECE Distinguished Alumni Award from UIUC in 2015, the Yaakov BarShalom Award for Lifetime Excellence in Information Fusion, ISIF in 2018, the Claude Shannon-Harry Nyquist Technical Achievement Award from the IEEE Signal Processing Society, the Pioneer Award from the IEEE Aerospace and Electronic Society in 2021, and Syracuse University Chancellor’s Lifetime Achievement Award in 2023.
Abstract : This talk will provide an introduction and overview of the field of optical signal processing, with a focus on a highly efficient general methodology using linear phase-only light-wave manipulations. This methodology has enabled the realization of many novel and greatly enhanced signal analysis and processing functionalities for a wide range of applications, from high-speed telecommunications to sensing and spectroscopy, using simple fiber-optics or integrated-waveguide device technologies. To illustrate the general methodology, the talk will provide an in-depth insight into a new framework of broad practical interest, namely, passive amplification of time- and frequency-domain waveforms with unique denoising capabilities for both classical signals and quantum correlation functions. This approach to noise mitigation enables the recovery of information that could not be accessed otherwise, pushing new frontiers in both fundamental and applied sciences.
Bio : José Azaña (Optica Fellow) received the Telecommunication Engineer degree and PhD degree in Telecommunication Engineering from the Universidad Politécnica de Madrid (UPM), Spain, in 1997 and 2001, respectively. Following research internships at the University of Toronto in Canada (1999) and the University of California – Davis in USA (2000), he conducted postdoctoral research work at McGill University in Montreal, Canada (2001-2003). Subsequently, he joined the Institut National de la Recherche Scientifique – Centre Energie, Matériaux et Télécommunications (INRS-EMT) in Montreal, where he is currently a Professor, and has been the holder of the Canada Research Chair in “Ultrafast Photonic Signal Processing”.
Prof. Azaña’s research interests include temporal optics, photonics-enabled broadband signal generation, characterization and processing, all-fiber and integrated-waveguide technologies, optical telecommunications, and quantum photonics. He has served in the technical program committee of numerous scientific conferences and technical meetings, and presently, he is a Senior Editor of the IEEE Photonics Journal and an Associate Editor of the IEEE Photonics Technology Letters. Prof. Azaña’s research outcome has been recognized with several research awards and distinctions, including the XXII national prize for the “best doctoral thesis in data networks” from the Association of Telecommunication Engineers of Spain (2002), the “extraordinary prize for the best doctoral thesis” from UPM (2003), the 2008 IEEE-Photonics Society Young Investigator Award, the 2009 IEEE-MTT Society Microwave Prize, and the 2020 Canada Brockhouse Prize for interdisciplinary research in science and engineering.
Bio : Radha Krishna Ganti is an Associate Professor at the Indian Institute of Technology Madras, Chennai, India. He received his B. Tech. and M. Tech. in EE from the Indian Institute of Technology, Madras, and a Masters in Applied Mathematics and a Ph.D. in EE from the University of Notre Dame in 2009. His doctoral work focused on the spatial analysis of interference networks using tools from stochastic geometry. He received the 2014 IEEE Stephen O. Rice Prize, and the 2014 IEEE Leonard G. Abraham Prize and the 2015 IEEE Communications society young author best paper award. He was also awarded the 2016-2017 Institute Research and Development Award (IRDA) by IIT Madras. In 2019, he was awarded the TSDSI fellow for technical excellence in standardisation activities and contribution to LMLC use case in ITU. He was the lead PI from IITM involved in the development of 5G base stations for the 5G testbed project funded by DoT.
Bio : Pranav Jha works with IIT Bombay, Mumbai, India as a Senior Scientist. His current research interest lies in Broadcast-Broadband Convergence, Rural Broadband Communication and Network Architecture for 5G & beyond. He is an active participant in the development of telecom standards and contributes to IEEE, 3GPP, TSDSI, and ITU standardization efforts. He also chairs two IEEE working groups, IEEE P1930.1 and IEEE P2061.
Abstract : This talk focuses on answering theoretical questions related to solving Federated Learning (FL) problems using a well known algorithm called Federated Average (FedAvg). FL is a distributed learning paradigm where multiple clients each having access to a local dataset collaborate with a server to solve a joint problem. The FedAvg algorithm is characterized by partial client participation and local updates at each client. Regardless of its popularity, the performance of FedAvg is not very well understood, especially in the interpolation regime, a common phenomenon observed in modern overparameterized neural networks such as deep neural networks. This talk addresses this challenge by performing a rigorous theoretical analysis of FedAvg for a class of non-convex functions satisfying the Polyak-Łojasiewicz (PL) inequality, a condition satisfied by overparameterized neural networks. For the first time, we establish that FedAvg with partial client participation achieves a linear convergence rate of O(log(1/ε)), where \epsilon is the solution accuracy. A decentralized version of this problem will also be discussed. Experiments on multiple real datasets corroborate our theoretical findings. If time permits, generealization capabilities of the FedAvg algorithm will be discussed.
This talk is mostly based on the extended version of the following publication:
Bio : B. N. Bharath completed his B.E. degree in Electrical and electronics from B. M. S. college of engineering, Bengaluru in 2005, and a direct Ph.D from the ECE Dept. of IISc, Bangalore in 2013. After completion of his Ph.D, he worked at Qualcomm Inc, Bangalore from 2013 July to 2014 August as a senior engineer. From, 2014 to 2017, he worked at PESIT Banglare south campus as a faculty. Since 2017, he has been working as an assistant professor in the Department of Electrical, Electronics and Communication engineering at IIT Dharwad. His research interests are in the broad area of Machine learning, distributed optimization, federated learning, and wireless communication/networks.
Abstract : We study the problem of distributed sampling and detection of remote point processes. A remote source, modelled as a homogeneous Poisson counting process (PCP) and with one of two possible intensity values, is observed at multiple remote observers in noise. The observers have a ‘sampling’ constraint which limits them from forwarding the entirety of their observations to a centralized fusion center, or ‘detector’. More precisely, the observers or samplers can remain active or ON for a fixed fraction of time over a fixed, known and finite time window. The detector seeks to design the sampler ON times for the distributed samplers jointly so as to optimize the overall accuracy of its detection procedure. We seek to understand this design of sampling-cum-detection strategies jointly. Our problem finds application in settings where the distributed agents communicating with a centralized fusion center are energy constrained or limited via finite capacity connecting links thereby limiting communication capability
Our main contribution is the complete characterization of optimal strategies for joint sampling and detection of the remote source. We first present optimal strategies for the two-sampler configuration, and then extend the characterization for the K-sampler, K > 2, case. Our results reveal a fundamental tension in the design of distributed sampling strategies between (i) overlapped distributed sampling for better noise rejection to obtain concurrent but independently corrupted source observations at multiple samplers, and (ii) disjointed distributed sampling for better source discovery. Our results also reveal an interesting fact that two simultaneously active samplers are necessary and sufficient for complete noise-rejection.
This is joint work with Vanlalruata Ralte (IIT Kharagpur), Stefano Rini (NYCU Taiwan).
**Please note that a part of this work is published and was presented at the IEEE International Symposium on Information Theory 2024.
Bio : Amitalok J. Budkuley is an assistant professor in the Dept. of Electronics and Electrical Communication Engineering at the Indian Institute of Technology Kharagpur since 2019. He received his B. Engg. degree in Electronics and Telecommunications Engineering from Goa University, in 2007, and his M. Tech. and Ph. D. degree in Electrical Engineering from the Indian Institute of Technology Bombay, Mumbai, India in 2009 and 2017 respectively. In between, he spent some time in industry working with Cisco Systems Inc.. From 2016 to 2019, he was at the Dept. of Information Engineering, The Chinese University of Hong Kong (CUHK) as a research assistant and then as a post-doctoral fellow.
His research interests include information theory, security and cryptography, signal processing for communication and control, and wireless communications.
Abstract : Data centers are becoming the backbone of digital economy as they support all sorts of applications. Furthermore, with recent advancements in AI/ML tools, including generative AI and its applications, the demand for increase in numbers and capacities of data centers is now growing explosively. By year 2030, it is projected that more than 8% of the global electricity demand would come from the data centers. Optical interconnects become crucial in the data centers as AI workloads specifically demand extremely high bandwidths for training and interference, among other applications.
A continued improvement in the capacity, cost, energy efficiency and latency is needed in the optical transceivers for such interconnects. This talk will present an overview of the underlying technologies and trends that support optical transceivers going from 800Gbps per module to several Tbps/module, which include direct detect and coherent techniques based pluggable optics modules, and co-packaged optics based solutions. The advancements made by our team in this area would also be presented.
Bio : Shalabh Gupta is currently a Professor of Electrical Engineering at IIT Bombay. Prior to joining IIT Bombay, he has worked in industry in the area of high-speed analog/RF integrated circuits and coherent optical links. His current research interests include optical interconnects, and high-speed electronic and photonic integrated circuits. Dr Gupta has been an inventor/co-inventor in more than 15 pending/granted US and Indian patents in the area. He is also a founder of the domestic startup Aortic Labs that is aimed at developing optical transceivers for high-capacity cost and energy efficient data center interconnects, using the path-breaking technologies developed indigenously by his group at IIT Bombay.
Abstract : Supervised learning, the harbinger of machine learning over the last decade, has had tremendous impact across application domains in recent years. However, the notion of a static trained machine learning model that identifies one among a few pre-specified set of classes is becoming increasingly limiting, as these models are deployed in changing and evolving environments. Among a few related settings, open-set and open-world learning have gained interest among practitioners to address this need of learning from new information, including the much-needed ability of saying "I don't know". In this talk, we will briefly discuss these settings, and highlight their importance in addressing real-world challenges. The talk will cover some of our recent research on open-world object detection (CVPR 2021), novel class discovery (ECCV 2022), and open-set object detection (WACV 2024) -- and also share interesting real-world use cases of these efforts. The talk will conclude with pointers to what could be ways to move forward as a community in this direction.
Bio : Vineeth N Balasubramanian is a Professor in the Department of Computer Science and Engineering at the Indian Institute of Technology, Hyderabad (IIT-H), India, and was recently a Fulbright-Nehru Visiting Faculty Fellow at Carnegie Mellon University until July 2023. He is also the Founding Head of the Department of Artificial Intelligence at IIT-H from 2019-22. His research interests include deep learning, machine learning, and computer vision with a focus on explainability, continual learning and learning with limited labeled data. His research has been published at premier venues including ICML, CVPR, NeurIPS, ICCV, KDD, AAAI, and IEEE TPAMI, with Best Paper Awards at recent venues such as CODS-COMAD 2022, CVPR 2021 Workshop on Causality in Vision, etc. He was recently the General Chair of ACML 2022 held in India, and regularly serves in senior roles for conferences such as CVPR, ICCV, AAAI, IJCAI, ECCV. He is listed among the World's Top 2% Scientists (2022,2023), and is a recipient of the Research Excellence Award at IIT-H (2022), Google Research Scholar Award (2021), NASSCOM AI Gamechanger Award (2022), Teaching Excellence Award at IIT-H (2017 and 2021), Outstanding Reviewer Award (IJCAI 2023, ICLR 2021, CVPR 2019, etc), among others. For more details, please see https://people.iith.ac.in/vineethnb.
Abstract : Integrated sensing and communications (ISAC) are envisioned to be an integral part of future wireless systems. In this talk, we will discuss model-drivel neural model to design transmit precoders for ISAC systems to jointly optimize a certain sensing and communications quality of service (QoS). In particular, we pose the problem of learning transmit precoders from uplink pilots and echoes as a function estimation problem and we model and parameterise this function using a neural model. To learn the neural network parameters, we develop a loss function based on the first-order optimality conditions to incorporate the sensing and communications QoS. Through numerical simulations, we demonstrate that the proposed method outperforms traditional optimization-based methods in presence of channel estimation errors while incurring lesser computational complexity and generalizing well across different channel conditions that were not shown during training.
Bio : Sundeep Prabhakar Chepuri received his M.Sc. degree (cum laude) in electrical engineering and Ph.D. degree (cum laude) from the Delft University of Technology, The Netherlands, in July 2011 and January 2016, respectively. He was a Postdoctoral researcher at the Delft University of Technology, The Netherlands. He has held positions at Robert Bosch, India, during 2007- 2009, and Holst Centre/imec-nl, The Netherlands, during 2010-2011. Currently, he is an Assistant Professor at the Department of ECE at the Indian Institute of Science (IISc) in Bengaluru, India.
Dr. Chepuri was a recipient the Pratiksha Trust Young Investigator award. His papers have received best paper awards at the ICASSP in 2015, ASILOMAR 2019, and EUSIPCO 2023, He was an Associate Editor of the EURASIP Journal on Advances in Signal Processing. Currently, he is an elected member of the EURASIP Technical Area Committee (TAC) on Signal Processing for Multisensor Systems, IEEE SPS Sensor Array and Multichannel Technical Committee (SAM-TC), IEEE SPS Signal Processing Theory and Methods Technical Committee (SPTM-TC), and an Associate Editor of IEEE Signal Processing Letters and IEEE Transactions on Signal and Information Processing over Networks. His general research interest lies in the field of mathematical signal processing, statistical inference, and machine learning applied to network sciences and wireless communications.
Abstract : This talk will provide an overview of two recent trends in computer networking, namely, Network Function Virtualization (NFV) and Software Defined Networking (SDN). The talk will then talk about how the ideas of NFV and SDN are being used in telecom networks like 5G networks today. We will discuss some recent research from our group that addresses the challenges in designing telecom networks using the principles of NFV and SDN. We will also provide some lessons on how to design future telecom networks to avoid such challenges.