News & Events
Beihang 70th Anniversary Series: High-end Academic Lectures of Institute of Artificial Intelligence
Release time:October 28, 2022

Time: Saturday, October 29, 2022, 08:00-17:00

Live Streaming Platform: Tencent Meeting

Conference ID: 363-372-961

Lecture 1: Robust Rank Aggregation and Its Application

Speaker: Prof. Ivor W. Tsang, IEEE Fellow, A*STAR, Singapore

Time Span: 8:00-9:00 a.m.


In rank aggregation (RA), a collection of preferences from different users are summarized into a total order under the assumption of homogeneity of users. Model misspecification in RA arises since the homogeneity assumption fails to be satisfied in the complex real-world situation. Existing robust RAs usually resort to an augmentation of the ranking model to account for additional noises, where the collected preferences can be treated as a noisy perturbation of idealized preferences. Since the majority of robust RAs rely on certain perturbation assumptions, they cannot generalize well to agnostic noise-corrupted preferences in the real world. In this talk, I first summarize the literature of Robust RA methods, and I present CoarsenRank, which possesses robustness against model misspecification. Specifically, the properties of our CoarsenRank are summarized as follows: (1) CoarsenRank is designed for mild model misspecification, which assumes there exist the ideal preferences (consistent with model assumption) that locate in a neighborhood of the actual preferences. (2) CoarsenRank then performs regular RAs over a neighborhood of the preferences instead of the original data set directly. Therefore, CoarsenRank enjoys robustness against model misspecification within a neighborhood. (3) The neighborhood of the data set is defined via their empirical data distributions. (4) CoarsenRank is further instantiated to Coarsened Thurstone, Coarsened Bradly-Terry, and Coarsened Plackett-Luce with three popular probability ranking models. Meanwhile, tractable optimization strategies are introduced with regards to each instantiation respectively. Finally, I present the applications of RAs in Neuroscience, Deep Generative Models, and Contrastive learning.

About the Speaker:

Prof Ivor W Tsang is Director of A*STAR Centre for Frontier AI Research (CFAR) since Jan 2022. Previously, he was a Professor of Artificial Intelligence, at University of Technology Sydney (UTS), and Research Director of the Australian Artificial Intelligence Institute (AAII), the largest AI institute in Australia, which is the key player to drive the University of Technology Sydney to rank 10th globally and 1st in Australia for AI research, in the latest AI Research Index. Prof Tsang is working at the forefront of big data analytics and Artificial Intelligence. His research focuses on transfer learning, deep generative models, learning with weakly supervision, big data analytics for data with extremely high dimensions in features, samples and labels. His work is recognised internationally for its outstanding contributions to those fields.

In 2013, Prof Tsang received his ARC Future Fellowship for his outstanding research on big data analytics and large-scale machine learning. In 2019, his Journal of Machine Learning Research paper titled "Towards ultrahigh dimensional feature selection for big data" received the International Consortium of Chinese Mathematicians Best Paper Award. In 2020, he was recognized as the AI 2000 AAAI/IJCAI Most Influential Scholar in Australia for his outstanding contributions to the field, between 2009 and 2019. His research on transfer learning was awarded the Best Student Paper Award at CVPR 2010 and the 2014 IEEE TMM Prize Paper Award. In addition, he received the IEEE TNN Outstanding 2004 Paper Award in 2007 for his innovative work on solving the inverse problem of non-linear representations. Recently, Prof Tsang was conferred the IEEE Fellow for his outstanding contributions to large-scale machine learning and transfer learning.

Besides these, Prof Tsang serves as the Editorial Board for the Journal of Machine Learning Research, Machine Learning, Journal of Artificial Intelligence Research, IEEE Transactions on Pattern Analysis and Machine Intelligence, IEEE Transactions on Artificial Intelligence, IEEE Transactions on Big Data, and IEEE Transactions on Emerging Topics in Computational Intelligence. He serves as a Senior Area Chair/Area Chair for NeurIPS, ICML, AAAI and IJCAI, and the steering committee of ACML.

Lecture 2: Predefined-Time Convergence and Its Application to Stabilization of Permanent-Magnet Synchronous Motor

Speaker: Prof. Michael Basin, Autonomous University of Nuevo Leon, Mexico

Time Span: 9:00-10:00 a.m.


First, a predefined-time convergent continuous linear time-varying control law is designed to stabilize a scalar model of the permanent-magnet synchronous motor system. Three cases have been considered: disturbance-free, in presence of a deterministic disturbance satisfying a Lipschitz condition, and in presence of both a stochastic white noise and a deterministic disturbance satisfying a Lipschitz condition. Numerical simulations are provided for a permanent-magnet synchronous motor system to validate the obtained theoretical results in each of the three considered cases. The simulation results demonstrate that the employed values of the predefined-time convergent control inputs are applicable in practice. Second, a predefined-time convergent continuous linear time-varying control law is designed, based on the backstepping technique, for a full-scale 4D permanent-magnet synchronous motor system with completely and incompletely measurable states, in presence of a deterministic disturbance satisfying a Lipschitz condition, and in presence of both a stochastic white noise and a deterministic disturbance satisfying a Lipschitz condition. The numerical simulations confirm the algorithm efficiency in each considered case. To the best of the authors’ knowledge, this is the first attempt to design a predefined-time convergent continuous control law for multi-dimensional systems with incompletely measurable states, using a linear time-varying control input.

About the Speaker:

Michael V. Basin received the Ph.D. degree in physical and mathematical sciences with major in automatic control and system analysis from Moscow Aviation University (MAI), in 1992. He is currently a Full Professor with the Autonomous University of Nuevo Leon, Mexico, and a Leading Researcher with ITMO University, St. Petersburg, Russia. Since 1992, he has published more than 300 research papers in international referred journals and conference proceedings. He has supervised 15 Ph.D. and nine master's theses. He is the author of the monograph New Trends in Optimal Filtering and Control for Polynomial and Time-Delay Systems (Springer). His works are cited more than 5,000 times (H index = 40). His research interests include optimal filtering and control problems, stochastic systems, time-delay systems, identification, sliding mode control, and variable structure systems. He is a Regular Member of the Mexican Academy of Sciences. He was awarded a title of a Highly Cited Researcher by Thomson Reuters, the publisher of Science Citation Index, in 2009. He has served as the Editorin-Chief and a Senior Editor in control for Journal of The Franklin Institute, a Technical Editor for IEEE/ASME Transactions on Mechatronics, and an Associate Editor for Automatica, IEEE Transactions on Systems, Man and Cybernetics: Systems, IET Control Theory and Applications, International Journal of Systems Science, and Neural Networks.

Lecture 3: Robust Cooperative Output Regulation of Heterogeneous Linear Multi-Agent Systems with Unbounded Transmission Delays

Speaker: Prof. Gang Feng, IEEE Fellow, City University of Hong Kong, China

Time Span: 10:00-11:00 a.m.


This talk presents our recent work on robust cooperative output regulation of heterogeneous uncertain linear multi-agent systems under unbounded distributed transmission delays. A novel distributed observer is first proposed to estimate the state of an exosystem in the presence of unbounded distributed transmission delays. Then two novel distributed controllers, one based on state feedback and the other based on output feedback, are further developed without any prior knowledge of the unbounded transmission delays. It is shown that the resulting closed loop multi-agent systems can achieve robust cooperative output regulation. It is also shown that better transient performance is achieved with our proposed controllers in contrast to many existing low gain controllers which are proposed to deal with both bounded and unbounded transmission delays. Our results can be directly applied to solve cooperative output regulation problems of multi-agent systems with bounded distributed or constant transmission delays. Furthermore, our results can also be directly applied to solve leader-following consensus problems of multi-agent systems with unbounded distributed, bounded distributed or constant transmission delays. Finally, a simulation example is provided to illustrate the effectiveness of the proposed controllers.

About the Speaker:

Gang Feng received the B.Eng and M.Eng. Degrees in Automatic Control from Nanjing Aeronautical Institute, China in 1982 and in 1984 respectively, and the Ph.D. degree in Electrical Engineering from the University of Melbourne, Australia in 1992.

Professor Feng was a Lecturer in Royal Melbourne Institute of Technology, 1991 and a Senior Lecturer/Lecturer, University of New South Wales, 1992-1999. He has been with City University of Hong Kong since 2000, where he is now a Chair Professor of Mechatronic Engineering. He has received ChangJiang Chair Professorship award, Alexander von Humboldt fellowship, the IEEE Computational Intelligence Society Fuzzy Systems Pioneer Award, the IEEE Transactions on Fuzzy Systems Outstanding Paper Award, the outstanding research award and President award of CityU, and several best conference paper awards. He is listed as a SCI highly cited researcher by Clarivate Analytics, 2016-2021. His research interests include intelligent systems and control, networked control systems, and multi-agent systems and control.

Professor Feng is a fellow of IEEE. He has been the Associate Editor of IEEE Trans. Automatic Control, IEEE Trans. on Fuzzy Systems, IEEE Trans. Systems, Man, & Cybernetics, Mechatronics, Journal of Systems Science & Complexity, Journal of Guidance, Navigation & Control, and Journal of Control Theory and Applications. He is also on the advisory board of Unmanned Systems.

Lecture 4: Multi-Robots Collaborative Mapping and Formation Keeping

Speaker: Prof. Danwei Wang, IEEE Fellow, Nanyang Technological University, Singapore

Time Span: 11:00-12:00 a.m.


Future robotics deployments will be in large quantities for efficiency by scaling. Applications will require multi-robots with capabilities such as collaborative mapping and moving in formation. Multi-robots collaboration require technologies beyond autonomous technologies for single robot. This talk discusses a few developments in our laboratory on collaborative mapping and formation keeping in unstructured complex and GPS-denied environment. The focuses are 1) multi-robots collaborative exploration in an efficiency manner with merged map, and 2) robust formation keeping. Examples and experiments are used to illustrate the concepts and points.

About the Speaker:

Danwei Wang, Academician of the Singapore Academy of Engineering and IEEE Fellow, was awarded the Alexander von Humboldt Fellowship in Germany and the first prize in 2017 Science and Technology in Shanghai. He is currently a Professor in the School of Electrical and Electronic Engineering at Nanyang Technological University. He received his Ph.D. degree from the University of Michigan, Ann Arbor, USA in 1989. He is the Editor-in-Chief of IEEE/RSJ IROS, and has published 5 English monographs, 7 book chapters, 9 patents and more than 500 international referred journal and conference papers. Prof. Wang is actively involved in research related to robotic manipulators and force control, advanced control system design, intelligent system, learning control, mobile robot, mobile robot and trajectory control, satellites formation flying and fault-tolerant attitude control, fault diagnosis and prognosis for complex systems, and traffic light control. His papers have been cited more than 8,800 times by Science Citation Index (SCI) papers and more than 15,000 times in Google Scholar as of February 2022.

Lecture 5: Intelligence Based Spatiotemporal Modeling and Learning for Multiscale Dynamic Systems

Speaker: Prof. Hanxiong Li, IEEE Fellow, City University of Hong Kong, China

Time Span: 14:00-15:00 p.m.


We live in a world of time and space. There are many dynamic processes of space-time coupling in various fields of industry, from temperature distribution, fluid injection, to flexible mechanical arm motion. Such processes are collectively referred to as distributed parameter systems (DPS). Since the constitutive equations of DPS are usually partial differential equations, it is very difficult to model and simulate its dynamics during online operation, let alone detect any abnormalities that occur in the process.

In off-line physical simulation (the so-called digital twin), the core issue is parameter optimization of the system. Due to the complexity of the physical model described by partial differential equations, traditional optimization methods cannot be applied directly. According to the experience of experts and the specific characteristics of the system, an intelligent method of multi-objective mixing with parameter sensitivity selection and multi-scale calibration is proposed, which can improve model tuning. The parameter optimization is always a challenge for spatiotemporal dynamic process.

Online performance prediction requires analytical models for fast computation. This requires the partial differential equation to be approximated in a group of ordinary differential equations, under space-time separation. Different learning methods are developed to enhance the performance under uncertainties, and many methods in AI field can be applied for temporal learning.

Spatial abnormality is difficult to detect and locate. Model and sensors are critical to successful detection and localization. Very few work have been reported internationally.

The research methods have been applied to the analysis and prediction of the temperature field performance of automotive lithium batteries, as well as the curing process in IC packaging industry.

About the Speaker:

Han-Xiong LI received his B.E. degree in aerospace engineering from the National University of Defence Technology, China, M.E. degree in electrical engineering from Delft University of Technology, Delft, The Netherlands, and Ph.D. degree in electrical engineering from the University of Auckland, Auckland, New Zealand.

Currently, he is a chair professor in the Department of Advanced Design & Systems Engineering, the City University of Hong Kong. Over the last thirty years, he has had opportunities to work in different fields, including military service, investment banking, industry, and academia. He has authored 2 books and about 20 patents, and published more than 250 SCI journal papers with h-index 55 (web of science). He has been rated as highly cited Chinese scholar by Elsevier since 2014, and featured as “Top 1000 Scientists” according to Top Scientists Ranking for Computer Science & Electronics prepared by Guide2Research since 2020. His current research interests are in system intelligence and control, intelligent modeling and learning, distributed parameter systems, smart sensing and IoT.

Dr. Li serves as Associate Editor of IEEE Transactions on Systems, Man & Cybernetics: system (2016- ), IEEE Transactions on Cybernetics (2002-2016), and IEEE Transactions on Industrial Electronics (2009-2015). He was awarded the Distinguished Young Scholar (overseas) by the China National Science Foundation in 2004, a Chang Jiang scholar by the Ministry of Education, China in 2006, and a scholar in China Thousand Talents Program in 2010. He serves as the distinguished expert for Hunan Government and China Federation of Returned Overseas. He is a fellow of the IEEE.

Lecture 6: An Alternative Paradigm of Fault Diagnosis in Dynamic Systems: Projection-based Methods and Applications

Speaker: Prof. Steven X. Ding, University of Duisburg-Essen, Germany

Time Span: 15:00-16:00 p.m.


In the era of big-data, data-driven and machine learning methods become the most dominant research domain in technical fault diagnosis. Nevertheless, observer-based fault diagnosis technique is still the major efficient tool applied to address fault diagnosis issues for dynamic and particularly automatic control systems. In spite of extensive studies over past five decades, there exist a number of open and challenging issues in the observer-based framework, These include optimal detection of parametric (multiplicative) faults, in particular in the presence of model uncertainties as well as in feedback control systems, and lack of convincing and mathematically well-established metrics to measure the distance between nominal and faulty operations for fault detection purpose or the distance between two different classes of faults towards fault isolation. Recently, it is proposed to apply the orthogonal projection technique, as an alternative paradigm, for fault diagnosis in dynamic systems. In this talk, a short introduction is given to (i) the projection-based fault diagnosis framework, (ii) comparison with observer-based fault diagnosis systems, and (iii) applications of projection[1]based fault diagnosis framework to fault detection in dynamic systems, control system performance monitoring and design of safe automatic control systems.

About the Speaker:

Prof. Steven X. Ding received his Ph.D. degree from the University of Duisburg, Germany in 1992 and was appointed tenured professor at the University of Applied Science Lausitz in Senftenberg, Germany in 1995. He then served there as Vice-President from 1998 to 2000. In 2001, he was appointed full-professor of Control Engineering at the University of Duisburg-Essen and Director of the Department of Automatic Control and Complex Systems. Since 2001, Prof Steven X. Ding has been working as coordinator and organiser of the EC research framework programmes (5th, 6th and 7th framework). Prof Steven X. Ding has long been a close cooperative partner with the German automotive and new energy industry.

Lecture 7: Understanding Multimodal Videos by Learning with Privileged Information and Distillation

Speaker: Prof. Vittorio Murino, IEEE Fellow, University of Verona, Italy

Time Span: 16:00-17:00 p.m.


Diverse input data modalities can provide complementary cues for several tasks, usually leading to more robust algorithms and better performance. However, while a (training) dataset could be accurately designed to include a variety of sensory inputs, it is often the case that not all modalities are available in real life (testing) scenarios, where a model has to be deployed. This raises the challenge of how to learn robust representations leveraging multimodal data in the training stage, while considering limitations at test time, such as noisy or missing modalities. In this talk, I will present a new approach for multimodal video action recognition, developed within the unified frameworks of distillation and privileged information, named generalized distillation. Particularly, we consider the case of learning representations from depth and RGB videos, while relying on RGB data only at test time. We propose a new approach to train a hallucination network that learns to distill depth features through multiplicative connections of spatio-temporal representations, leveraging soft labels and hard labels, as well as distance between feature maps. Subsequently, we improve the hallucination model to distill depth information via adversarial learning, resulting in a clean approach without several losses to balance or hyperparameters to tune. We report state-of-the-art results on video action classification on multimodal datasets such as NTU RGB+D, UWA3DII and Northwestern-UCLA.

About the Speaker:

Vittorio Murino is full professor at the University of Verona, Italy, and visiting scientist of PAVIS (Pattern Analysis and Computer Vision) department at the Istituto Italiano di Tecnologia, in Genova, Italy. He took the Laurea degree in Electronic Engineering in 1989 and the Ph.D. in Electronic Engineering and Computer Science in 1993 at the University of Genova, Italy. He was chairman of the Department of Computer Science from 2001, year of foundation, to 2007, and coordinator of the Ph.D. program in Computer Science in the same university from 1999 to 2003. From 2009 to 2019, he worked at the Istituto Italiano di Tecnologia in Genova, Italy, as director of the PAVIS (Pattern Analysis and Computer Vision) department. From 2019 to 2021, he worked as Senior Video Intelligence Expert at the Ireland Research Centre of Huawei Technologies (Ireland) Co., Ltd. in Dublin. His main research interests include computer vision, pattern recognition and machine learning, nowadays focusing on deep learning approaches, domain adaptation and generalization, and multimodal learning for (human) behavior analysis and related applications, such as video surveillance and biomedical imaging. Prof. Murino is co-author of more than 400 papers published in refereed journals and international conferences, member of the technical committees of important conferences (CVPR, ICCV, ECCV, ICPR, ICIP, etc.), and guest coeditor of special issues in relevant scientific journals. He is also member of the editorial board of Computer Vision and Image Understanding and Machine Vision and Applications journals. Finally, prof. Murino is IEEE Fellow, IAPR Fellow, and ELLIS Fellow.

Institute of Artificial Intelligence, Beihang University