Member of the US National Academy of Engineering, IEEE Fellow
Carnegie Mellon University
Learning in the Real World: Diverse, Spatial, Temporal Dependent Data
In the real world, data comes from many different sources (e.g., multiple devices), with inter-dependencies (across sources, aka spatial dependencies) and intra-dependencies (across time), multiple types, and different nature (e.g., weather patterns, traffic counts). Deep Learning (DL) has shown significant improvements in computer vision and natural language with ever larger and more sophisticated deep models. But in many other application domains, annotated data is scarce, data dependencies are of paramount significance, and often there is prior knowledge and there are extensive libraries of data processing methods and algorithms that have served us well in the past. In this lecture, we consider some of these issues – adopting graphs to capture data spatial dependencies, and strategies to learn from such complicated diverse data. We illustrate our results with benchmark datasets and real-world applications (e.g., traffic flows and taxi requests in urban environments).
José M. F. Moura is the Philip L. and Marsha Dowd University Professor at CMU, with interests in signal processing and data science. A detector in two of his patents with Alek Kavcic is found in over 60% of the disk drives of all computers sold worldwide in the last 15 years (4 billion and counting)–leading to a US $750 Million settlement between CMU and Marvell. He was the 2019 President and CEO of IEEE, the largest professional society in the world with its 420 thousand members. He was Editor in Chief for the Transactions on SP. He is Fellow of the IEEE, AAAS, and the US National Academy of Inventors, holds honorary doctorate degrees from the University of Strathclyde (UK) and Universidade de Lisboa (Portugal), he is a member of the Academy of Sciences of Portugal, and a member of the US National Academy of Engineering. He received the Great Cross of the Order of The Infante D. Henrique bestowed to him by the President of the Republic of Portugal.
Zhi-Quan (Tom) Luo
Foreign Member of the Chinese Academy of Engineering, IEEE Fellow
The Chinese University of Hong Kong, Shenzhen
Rethinking WMMSE in the Era of Massive MIMO
Precoding design for maximizing weighted sumrate (WSR) is a fundamental problem for downlink of massive multi-user multiple-input multiple-output (MU-MIMO) systems.
It is well-known that this problem is generally NP-hard due to the presence of multi-user interference. The weighted minimum mean-square error (WMMSE) algorithm is a popular approach for WSR maximization.
However, its computational complexity is cubic in the number of base station (BS) antennas, which is unaffordable when the BS is equipped with an extremely large antenna array.
In this talk, we will reshape the classical WMMSE with theoretical guarantee and linear complexity (i.e., scales linearly with the number of BS antennas).
Moreover, we show that the novel WMMSE algorithm can be also used to reduce the communication overhead for coordinated multi-point transmission.
Zhi-Quan (Tom) Luo is the Vice President (Academic) at The Chinese University of Hong Kong, Shenzhen where he has been a professor since 2014.
He is the Director of Shenzhen Research Institute of Big Data and also the Chinese University of Hong Kong (Shenzhen)-Shenzhen Research Institute of Big Data-Huawei Innovation Laboratory of Future Network System Optimization.
He completed his Ph.D. at the Massachusetts Institute of Technology and his undergraduate studies at Peking University, China.
His research interests lie in the area of big data, signal processing and digital communication, ranging from theory to design to implementation.
He has served on more than forty conference and workshop program committees and been the Chair of the IEEE Signal Processing Society Technical Committee on Signal Processing for Communications (SPCOM).
He was the Editor in Chief for IEEE Transactions on Signal Processing from 2012 to 2014 and served as the Associate Editor for many internationally recognized journals.
Currently, he is the Fellow of the Institute of Electrical and Electronics Engineers (IEEE) and the Society for Industrial and Applied Mathematics (SIAM).
He was elected to Foreign Member of the Chinese Academy of Engineering (CAE) in 2021.
He received the 2010 Farkas Prize from the INFORMS Optimization Society.
He also received three Best Paper Awards from the IEEE Signal Processing Society in 2004, 2009 and 2011 respectively, and a 2011 Best Paper Award from the EURASIP.
In 2014, he was elected to the Royal Society of Canada. In 2018, he was awarded the prize of Paul Y. Tseng Memorial Lectureship in Continuous Optimization.
Converging technological advances in sensing, machine learning and computing offer tremendous opportunities for continuous contextually rich yet unobtrusive multimodal, spatiotemporal characterization of an individual’s behavior and state, and of the environment within which they operate. This in turn is enabling novel possibilities for understanding and supporting various aspects of human-centered applications notably in psychological health and well-being. This talk will highlight some of the advances, opportunities and challenges in gathering human-focused data and creating algorithms for machine processing of such cues. It will report efforts in Behavioral Signal Processing (BSP)—technology and algorithms for quantitatively and objectively understanding typical, atypical and distressed human behavior—with a specific focus on communicative, affective and social behavior.
Examples will be drawn from health and wellbeing realms such as Autism Spectrum Disorder, Couple therapy, Depression, Suicidality, and work place behavior. It will also discuss the challenges and opportunities in creating trustworthy signal processing and machine learning approaches that are inclusive, equitable, robust, safe and secure e.g., with respect to protected variables such as gender/race/age/ability etc.
Shrikanth (Shri) Narayanan is University Professor and Niki & C.L. Max Nikias Chair in Engineering at the University of Southern California,
where he is Professor of Electrical & Computer Engineering, Computer Science, Linguistics, Psychology, Neuroscience, Pediatrics,
and Otolaryngology—Head & Neck Surgery, Director of the Ming Hsieh Institute and Research Director of the Information Sciences Institute.
Prior to USC he was with AT&T Bell Labs and AT&T Research. His research focuses on human-centered information processing and communication technologies.
He is a Guggenheim Fellow and a Fellow of the National Academy of Inventors, the Acoustical Society of America, IEEE, ISCA, the American Association for the Advancement of Science (AAAS),
the Association for Psychological Science, and the American Institute for Medical and Biological Engineering (AIMBE). He is a recipient of several honors including the 2015 Engineers Council’s Distinguished Educator Award,
a Mellon award for mentoring excellence, the 2005 and 2009 Best Journal Paper awards from the IEEE Signal Processing Society and serving as its Distinguished Lecturer for 2010-11,
a 2018 ISCA CSL Best Journal Paper award, and serving as an ISCA Distinguished Lecturer for 2015-16, Willard R. Zemlin Memorial Lecturer for ASHA in 2017, and the Ten Year Technical Impact Award in 2014 and the Sustained Accomplishment Award in 2020 from ACM ICMI.
He has published over 900 papers and has been granted eighteen U.S. patents. His research and inventions have led to technology commercialization including through startups he co-founded: Behavioral Signals Technologies focused on the telecommunication services and AI based conversational assistance industry and Lyssn focused on mental health care diversity, treatment and quality assurance. He serves as the inaugural Vice President–Education for the IEEE Signal Processing Society.
Fellow of the Canadian Academy of Engineering, IEEE Fellow
University of Waterloo
Puzzles in Perceptual Image Quality Assessment and Processing
Perceptual image quality assessment and perceptually motivated processing have drawn a great deal of attention in the past two decades.
While a large number of methods have been developed for a wide variety of applications, somewhat surprisingly, limited thoughts have been put on accurately formulating the problem and clearly defining the performance evaluation criterion.
In this talk, we would like to share some of the puzzles we have for open discussion. For example, classical image processing problems such as image restoration and image enhancement are often evaluated and compared either by some quality metric of the reconstructed images or by certain signal fidelity/distortion measure between the original and reconstructed images, but we argue that neither of them is precisely the desired target for image restoration or enhancement.
While some compromise between the two might be a better option, there is a lack of clean theory or methodology to determine the balancing point.
Although a Bayesian interpretation may offer useful insights and potential solutions, the instantiations are often constrained by specific distortion processes.
In addition, whether the prior or posterior probability of images produces an appropriate representation of perceptual image quality is yet another question up to debate.
Recently, there has been some interesting discussion on the idea of the “perception-distortion tradeoff”, aiming to find some theoretical compromise between quality (perception) and distortion, but the quality (perception) is defined by the divergence between the distributions of the original and reconstructed images, which makes the problem even more perplexing.
We hope this talk can invite some insightful discussions, which may help with our future effort on perceptual image quality assessment and processing.
Zhou Wang is a Canada Research Chair and Professor in the Department of Electrical and Computer Engineering, University of Waterloo.
His research interests include image/video/multimedia processing, coding, communication, computational vision, and machine learning, with focuses on perceptual quality assessment and perceptually motivated processing.
He has more than 200 publications in the fields with over 80,000 citations based on Google Scholar statistics.
Dr. Wang is a Fellow of IEEE, a Fellow of Royal Society of Canada - Academy of Science, and a Fellow of Canadian Academy of Engineering.
He is a recipient of Steacie Memorial Fellowship awarded by the Governor General of Canada, and IEEE Signal Processing Society Best Paper Award, Best Magazine Paper Award, and Sustained Impact Paper Award.
He is also a two-time recipient of Engineering/Technology Emmy Awards – the most prestigious technology honor of the TV industry – one in 2015 as an individual, and the other in 2021 by SSIMWAVE Inc. of which he is a co-founder and the Chief Scientist.
Gordon Mckay Professor
Exploring and Exploiting the Universality Phenomena in High-Dimensional Estimation and Learning
Universality is a fascinating and truly high-dimensional phenomenon. It points to the existence of universal laws that govern the macroscopic behavior of wide classes of large and complex systems, despite their differences in microscopic details.
The notion of universality originated in statistical mechanics, especially in the study of phase transitions. Similar phenomena have been observed in probability theory, dynamical systems, random matrix theory, and number theory.
In this talk, I will present some recent progresses in rigorously understanding and exploiting the universality phenomena in the context of statistical estimation and learning on high-dimensional data. Examples include spectral methods for high-dimensional projection pursuit, statistical learning based on kernel and random feature models, and approximate message passing algorithms on structured and nearly deterministic data matrices. Together, they demonstrate the robustness and wide applicability of the universality phenomena.
Yue M. Lu was born in Shanghai.
After finishing undergraduate studies at Shanghai Jiao Tong University, he attended the University of Illinois at Urbana-Champaign, where he received the M.Sc. degree in mathematics and the Ph.D. degree in electrical engineering, both in 2007.
He is currently Gordon McKay Professor of Electrical Engineering and of Applied Mathematics at Harvard University.
He is also fortunate to have held visiting appointments at Duke University in 2016 and at the École Normale Supérieure (ENS) in 2019.
His research interests include the mathematical foundations of statistical signal processing and machine learning in high dimensions.
Professor, Chair of Electrical and Electronics Engineering Department
Machine Learning based Semantic Signal Processing and Communications
Recent advances in machine learning enabled real time extraction of semantic information in sensor data. By taking advantage of this development, a novel framework for goal oriented semantic signal processing and communications is proposed. Unlike classical approaches where sensor data is encoded and transmitted to a processing unit in the network, in the proposed framework, semantic information in the sensor data is extracted in real time by using machine learning techniques at the sensor node. To enable efficient goal-oriented signal processing on the extracted semantic information, a hierarchical graph-based semantic language is used. In this way, semantic filtering of the extracted information can be achieved at the sensor node resulting in a dramatic reduction in the rate of communication in the network. Effectiveness of the proposed approach is demonstrated over real video data obtained in a parking lot and in a simulated smart city environment.
Orhan Arikan was born in 1964, in Manisa, Turkey. In 1986, he received the B.Sc. degree in Electrical and Electronics Engineering from the Middle East Technical University, Ankara, Turkey. He received both the M.S. and Ph.D. degrees in Electrical and Computer Engineering from the University of Illinois Urbana-Champaign, in 1988 and 1990, respectively.
Following his graduate studies, he worked for three years as a Research Scientist at Schlumberger-Doll Research Center, Ridgefield, CT, USA. During this time, he was involved in the inverse problems and fusion of multiple modality measurements. In1993, he joined the Electrical and Electronics Engineering Department of Bilkent University, Ankara, Turkey. His research interests are in the areas of semantic signal processing, statistical signal processing and remote sensing. He served as the chairman of the department in 2011-2019. Currently, he is the Dean of Engineering Faculty at Bilkent University.
In 1998, He received the Distinguished Teaching Award of Bilkent University. In 2002, He received the Young Investigator Award in Engineering from Turkish Scientific and Technical Research Foundation. He has served as the Chairman of IEEE Signal Processing Society, Turkey Section in 1995-1996 and served as the President of IEEE Turkey Section in 2000-2001.
Microsoft Research Asia
Convergence of Computer Vision and Natural Language Processing
As mankind, we can accomplish various intelligence capabilities, such as vision, language, and science, simply by using a single neuron organ called the cerebral cortex. The pre-training of cortical nerons for different capabilities also relies heavily on a similar mechanism of predictive learning. These unified biological mechanisms have enabled human beings to adapt quickly and effectively to new environments and acquire new capabilities without millions of years' biological evolution.
In artificial intelligence, architectures and learning methods in various domains are also converging. Transformer, which emerges in the field of NLP, is now taking over previous domain-specific architectures in several fields, such as computer vision, speech, science, etc. Masked signal modeling or predictive learning has also been shown to be very effective in all of NLP, vision, and speech. This talk will introduce the journey towards these convergences, as well as the representative works that have driven this trend. The talk will also present several recent research efforts by the speaker's team, including Local Relation Networks, Swin Transformer V1/V2, SimMIM, etc.
Han Hu is currently a principal researcher in Microsoft Research Asia (MSRA), and an adjunct ph. D. advisor at Xi'an Jiaotong University. His main research interests include visual architecture design, self-supervised representation learning, and visual-language representation learning. His paper Swin Transformer won the ICCV 2021 Best Paper Award (Marr Prize) and is now widely used in academia and industry. He received his Ph. D. and bachelor's degrees from Tsinghua University in 2014 and 2008, respectively. His Ph. D. dissertation received Excellent Doctoral Dissertation Award of Chinese Association of Artificial Intelligence (CAAI). He visited the University of Pennsylvania for half a year in 2012 and worked in Institute of Deep Learning (IDL), Baidu Research, between 2014 and 2016. He served as an area chair for CVPR2021 and CVPR2022.
Assistant Professor, Presidential Young Fellow
The Chinese University of Hong Kong, Shenzhen
Learning for the Future Power Grid
Advanced learning frameworks are reshaping the landscape of power grid operation and the electricity market design.
This talk shares two stories, both of which seek to use learning frameworks to enhance the future power grid.
The first one investigates the storage control problem for consumers. Specifically, we consider that consumers face dynamic electricity prices and seek to use storage to reduce their electricity bills.
The challenges come from the uncertainty in the electricity price and consumers' demand. We propose a practical learning-based online storage control policy.
The second story studies a classical procedure in the electricity market, the economic dispatch problem, i.e., matching the electricity supply and demand at the minimal generation cost.
The critical challenge is again from the uncertainty in the system demand. Hence, the conventional approach is to conduct the dispatch based on predicted demand.
However, we submit that this conventional approach can be suboptimal, and we propose a model-free algorithm for economic dispatch based on the end-to-end learning framework.
Dr. Chenye Wu is currently an Assitant Professor and the presidential young fellow at the School of Science and Engineering.
The Chinese University of Hong Kong, Shenzhen. Dr. Wu received his bachelor's degree in electronic engineering from Tsinghua University in 2009 and his Ph.D. degree in computer science and engineering from Tsinghua University in 2013, advised by Prof. Andrew Yao,
the Turing Award Laurant. Dr. Wu's research interests span from power system control to the electricity market design, emphasizing the emerging business model design for the energy sector, the market power analysis for the electricity market, the AI-driven power system control and operation.
Dr. Wu has published over 70 research articles in top journals and leading conferences in the field, including IEEE Transactions on Power Systems, IEEE Transactions on Smart Grid, IEEE Transactions on Sustainable Energy, ACM e-Energy.
He is a member of the FinTech special interest group, China Society for Industrial and Applied Mathematics, and a member of the special interest group, China Energy Society. Dr. Wu has been an Editorial Board Member for IEEE Systems Journal as an Associate Editor since February 2022.
He is the symposium co-Chair for IEEE SmartGridComm 2022 and the digital conference co-Chair for ACM e-Energy 2022. Dr. Wu is the co-recipients of the three best paper awards, including the best paper award for IEEE SmartGridComm 2012 and IEEE PES General Meeting 2013 and 2020.
Shanghai Jiao Tong University
Bridging Graph Signal Processing and Graph Neural Networks
Data today is often generated from a diverse sources, including social, citation, biological, and physical infrastructure. Unlike time-series signals or images, such signals possess complex and irregular structures, which can be modeled as graphs. Analyzing graph signals requires dealing with the underlying irregular relationships. To achieve this, graph convolutional networks and variants permeats the benefits of deep learning to the graph domain, achieving remarkable success in social network analysis, traffic analysis and quantum chemistry. However, many network architectures are designed through trial and error. It is thus hard to explain the design rationale and further improve the architectures.
Siheng Chen is a tenure-track associate professor of Shanghai Jiao Tong University.
Before joining Shanghai Jiao Tong University, he was a research scientist at Mitsubishi Electric Research Laboratories (MERL), and an autonomy engineer at Uber Advanced Technologies Group (ATG), working on the perception and prediction systems of self-driving cars. Before joining industry,
Dr. Chen was a postdoctoral research associate at Carnegie Mellon University. Dr. Chen received his doctorate in Electrical and Computer Engineering from Carnegie Mellon University in 2016, where he also received two master degrees in Electrical and Computer Engineering (College of Engineering) and Machine Learning (School of Computer Science), respectively. He received his bachelor’s degree in Electronics Engineering in 2011 from Beijing Institute of Technology, China. Dr. Chen's work on sampling theory of graph data received the 2018 IEEE Signal Processing Society Young Author Best Paper Award. His co-authored paper on structural health monitoring received ASME SHM/NDE 2020 Best Journal Paper Runner-Up Award and another paper on 3D point cloud processing received the Best Student Paper Award at 2018 IEEE Global Conference on Signal and Information Processing. Dr. Chen contributed to the project of scene-aware interaction, winning MERL President's Award. His research interests include graph signal processing, graph neural networks and group intelligence.
Concurrent Order Dispatch for Instant Delivery with Time-Constrained Actor-Critic Reinforcement Learning
In this talk, I will first introduce some key concepts of data-driven cyber-physical systems in smart cities including (i) mobile sensing and communication based on data collected from IoT devices, (ii) data-fusion and cross-domain prediction,
and (iii) intelligent decision-making.
Then, I will introduce our group’s work on the applications of Human Cyber-Physical Systems in smart cities with a concrete use case, i.e.,
concurrent order dispatch for instant delivery with strict deadlines.
Thanks to the rapid development of online digital platforms, instant delivery services have developed rapidly in recent years and significantly changed the lifestyle of people.
In instant delivery, the delivery process needs to be finished in a short time. Given the massive number of orders and the limited number of couriers, it is essential to make appropriate dispatch decisions to improve delivery efficiency.
Based on a large number of historical delivery records, I will provide some data-driven findings and unique challenges of order dispatch in instant delivery.
A time-constrained reinforcement learning-based order dispatch framework is proposed, which models concurrent order dispatch as a sequential decision-making problem and optimizes dispatch decisions with actor-critic reinforcement learning.
Finally, I will discuss a few new challenges and open problems in on-demand delivery and intelligent logistics.
Dr. Shuai Wang is currently a professor with the School of Computer Science and Engineering, Southeast University, Nanjing, China.
He received the B.S. and M.S. degrees from the Huazhong University of Science and Technology, China, in 2009 and 2012, respectively, and the PhD degree in the Department of Computer Science and Engineering at the University of Minnesota in 2017.
His research interests include Data Science, Internet of Things, and Cyber-Physical Systems. Shuai has published more than 50 papers in premium conferences/journals, e.g., SIGKDD, ICDE, RTSS, MOBICOM, UBICOMP, INFOCOM, ICNP, TON, TPDS, TOSN, and TWC. He won the outstanding paper award of RTSS’21.