Digital Twin Network (DTN): The Third-generation Internet Application Platform
09:15-09:35,October 9,2021

Speaker: Qinping Zhao, Academician ,Chinese Academy of Engineering,China


The connected object of the first-generation Internet is computer, and its applications are primarily social communication and marketing; The secondgeneration Internet is the Internet of things (IoT), which has widely applied in various industries and living services. The application of digital twin expands the connected objects of the IoT into physical objects and their virtual twins, thus integrates the physical and virtual object space into a virtual-real hybrid space —Digital Twin Network(DTN) . Developing the Internet of Things into a new generation of DTN can greatly improve the efficiency of production and service operations in various industries.

We should actively develop the DTN to provide interconnected support for the in-depth VR applications in various physical industries. The new generation of digital twin networks, combined with software and hardware VR infrastructure such as cloud VR platform, VR data center, VR model center, VR engine and content production platform, can comprehensively support the VR industry chain and construction and formation of the industry ecology, thereby promoting the rapid and healthy development of the VR industry .

Short Bio:

Qinping Zhao, the professor of Beihang University, is also a member of the China Academy of Engineering. Deeply engaged in virtual reality (VR) and artificial intelligence ering.esearch, he presided over 20 national science and technology programs, like the National Natural Science Foundation of China, the National 863 Program, and the National Defense Pre-research programs. The real-time 3D graphics platform he developed won the first prize of State Scientific and Technological Progress Award in 2010. The virtual scene generation technology driven by video image content won the second prize of State Technological Invention Award in 2014. He has also developed the virtual reality application systems such as tactical command simulation training system, design evaluation system of aircraft cockpit based on mixed reality, the creative simulation and process monitoring system for the Beijing Olympic Games Opening ceremony, and 3D reasoning and decision-making system for the 60th Anniversary of the National Day Parade. He has published more than 200 academic papers, with over 60 national invention patents, and 4 monographs.

Unlearnable examples, adversarial examples and noisy label examples: testing the robustness of deep neural networks
10:45-11:30,October 9,2021

Speaker: James Bailey, Professor, University of Melbourne, Australia


Deep neural networks (DNNs) are a major force in machine learning and have achieved great success across many domains. However, DNNs may behave in highly non robust ways when presented with perturbed examples at either training or test time. In this talk we explore 3 important types of perturbations: adversarial, noisy label and unlearnable.

Adversarial examples are well known to the research community and correspond to specially perturbed inputs that fool DNNs into making false predictions with high confidence. Noisy label examples arise when a significant fraction of class labels in the training is incorrect, making it challenging to fit an accurate model. Unlearnable examples are a more recent phenomenon and correspond to training data that undergoes small intentional perturbations that make it useless for model training, whilst preserving utility for human visualization.

In this talk we compare and contrast these three types of perturbations, examine their impact on the robustness of DNNs, mitigation strategies, and consider their broader implications for robust machine learning.

Short Bio:

Professor James Bailey is a Professor in the Faculty of Engineering and Information Technology at The University of Melbourne and Program Lead for Artificial Intelligence. His research interests particularly relate to the assurance, certification and safety of systems based on machine learning and artificial intelligence. He contributes to the AI and data science communities through roles such as membership of Editorial Boards including the Journal of Artificial Intelligence Research, ACM Transactions on Data Science and IEEE Transactions on Big Data. He is co-PC Chair of the 2021 IEEE International Conference on Data Mining. He works on the development and deployment of AI and machine learning systems in collaboration with a wide range of industry and government partners across sectors including education, energy and health.

Blockchain as a platform for secure collaboration
14:30-15:15,October 9,2021

Speaker:  Elena Ferrari, Professor, University of Insubria, Italy


Today many applications requiring to analyze and process a massive amount of data (e.g. scientific workflows) rely on business processes that span different organizations. This may pose serious security and privacy threats to the data each organisation exposes during the collaboration. The weak trust relationships that may hold among the collaborating parties result in a potential lack of trust in how data/operations are managed by the different business partners. The need for a trustworthy way to support composite service certification becomes crucial to guarantee efficiency and transparency. A promising approach to tackle this issue is leveraging on blockchain. This brings the benefits of trust decentralization, transparency, and accountability of the service composition and management process. The talk will describe a blockchain based framework we have developed to deal with the privacy and security challenges in executing inter-organisational business processes on the blockchain.

Short Bio:

Elena Ferrari is a professor of Computer Science at the University of Insubria (Italy), where she leads the STRICT SociaLab. She received her Ph.D and M.Sc. degree in Computer Science from the University of Milano (Italy). Her research interests are in the broad area of cybersecurity, privacy, and trust. Current research includes security and privacy for Big Data and Iot, access control, machine learning for cybersecurity, risk analysis, blockchain, and secure social media. Prof. Ferrari has published several books and more than 240 papers in international journals and conference proceedings. She has received several awards, including the 2009 IEEE Technical Achievement Award, the ACM CODASPY Research Award, the ACM SACMAT 10 Year Test of Time Award, an IBM Faculty Award, and a Google Research Award. She is an ACM and IEEE Fellow. In 2018, she has been named one of the 50 most influential Italian women in tech.

Event Cube for Event Discovery, Analysis and Management: a Case Study on Suicidal Events
16:15-17:00,October 9,2021

Speaker: Qing Li, Dept of Computing, The Hong Kong Polytechnic University, Hung Hom, Kowloon, HKSAR, CHINA


The publicly available data, such as the massive and dynamically updated news and social media data streams (a.k.a. big data), covers a wide range of social activities, personal views, and expressions, emphasizing the importance of comprehending and discovering the knowledge patterns underlying big data. Establishing methodologies and techniques for discovering real-world events from such large amounts of data, as well as for analyzing and managing such events in an efficient and aesthetic manner, is crucial and challenging. In this talk, I will present an event cube (EC) framework devised to support various data collection, consolidation, fusion and detection, and analysis and management tasks for various events. More specifically, we present a mechanism for data collection over multiple data sources in both passive and active manners, and promote the mappings constructed from various representation spaces for data consolidation. Furthermore, multimodal fusion is devised to integrate multiple data intrinsic structures and learn discriminative data representations so as to process heterogeneous multimodal data efficiently. We also present an approach to detect a specific type of events, namely, suicidal events based on a BERT-LSTM model, resulting in a mechanism called BLAM which combines BERT with Bi-LSTM to extract deep and rich features, with emotion classification being utilized as an auxiliary task to perform multi-task learning, so as to identify suicidal ideation and triggering events. Finally, an EC model is devised to support event organization and contextualization via hierarchical and analytical operations. A case study will be provided to demonstrate the capabilities and benefits of the EC facilities supporting on-line analytical processing of suicidal events and their relationships.

Bio Sketch:

Qing Li received the B.Eng. degree in computer science from the Hunan University (Changsha), China, and the M.Sc. and Ph.D. degrees in computer science from the University of Southern California (Los Angeles), USA. He is a Chair Professor and the Head of Department of Computing, the Hong Kong Polytechnic University. His current research interests include multimodal data management, conceptual data modeling, social media computing, Web services, and e-learning systems. He has authored/co-authored over 480 publications in the above areas. Prof. Li is also actively involved in the research community and has served as an associate editor of several technical journals, including IEEE Transactions on Artificial Intelligence, IEEE Transactions on Cognitive and Developmental Systems (TCDS), IEEE Transactions on Knowledge and Data Engineering (TKDE), the ACM Transactions on Internet Technology (TOIT), Data Science and Engineering, ​and World Wide Web, in addition to being a Conference and Program Chair/Co-Chair of numerous major international conferences. He is a fellow of IEE/IET (UK), a senior member of IEEE, and a distinguished member of CCF (China). He has sat/been sitting in the Steering Committees of ACM RecSys, IEEE U-MEDIA, DASFAA, WISE, FFD, and ICWL.