Prof. Liang Hu

Dept. of Computer Science and Technology, Tongji University, China

Biography: Prof. Liang Hu is is a full professor with Tongji University and also the Chief AI Scientist with DeepBlue Academy of Science, China. His research interests include recommender systems, machine learning, data science and general intelligence. He has published a number of papers in top-rank international conferences and journals, including WWW, IJCAI, AAAI, ICDM, ICWS, TOIS, IEEE-IS. He has been invited as the program committee of more than 30 top-rank AI international conferences, including AAAI, IJCAI, ICDM, CIKM, and KDD. He also serves as the reviewer of more than ten AI and data science-related international journals, including ACM CSUR, IEEE TKDE, ACM TOIS, IEEE TPAMI, etc. In addition, he has presented eight tutorials on recommender systems and machine learning at top-rank AI conferences including IJCAI, AAAI, SIGIR, and ICDM.

Title: Federated Learning and Its Promising Applications.

Abstract: In recent years, every country has paid increasing attention to data security and personal privacy, which has a direct impact on socioeconomic development. Federated learning is a promising AI method that enables machine learning models to obtain knowledge from different datasets located on different devices or sites without sharing training data. This allows personal data to remain on local devices or sites, reducing the possibility of privacy breaches. In this talk, the speaker will introduce the basic concepts of federated learning and the classification of various federated learning approaches. After that, two real applications that demand high-level privacy will be introduced, where the federated learning methods are employed to guarantee the data and personal privacy. Firstly, the speaker will present the federated learning on crowdsourced HD mapping. Secondly, the speaker will present the potential application of federated learning on person re-identification.

Assoc. Prof. Chuanjun Zhao

School of Information, Shanxi University of Finance and Economics

Biography: Zhao Chuanjun, Associate professor of School of Information, Shanxi University of Finance and Economics, postdoctoral fellow of Computer Science and Technology, master tutor of Computer Application Technology, member of Emotional Computing Committee of Chinese Information Society of China. His research interests include data mining and natural language processing. In the early stage, it undertook the National Natural Science Foundation of China, the Industry-University Cooperative Education Project of the Ministry of Education, the Natural Science Foundation of Shanxi Province, the Science and Technology innovation Project of the Department of Education of Shanxi Province, the Graduate Education reform innovation project of Shanxi Province, and the Teaching reform innovation project of higher education institutions of Shanxi Province. In journals such as Information Science, Computer Speech and Language, Knowledge-based Systems, Computer Research and Development, Software Journal, Proceedings of the 2014 IEEE International Conference on Data Mining Workshop (ICDMW), Proceedings of the The Chinese National Conference on Social Media Processing has published more than 10 high-level papers, applied for and obtained a number of invention patents, and approved more than 30 software Copyrights.

Title:Cross-domain sentiment classification

Abstract:Training data in a specific domain are often insufficient in the area of text sentiment classifications. Cross-domain sentiment classification (CDSC) is usually utilized to extend the application scope of transfer learning in text-based social media and effectively solve the problem of insufficient data marking in specific domains. Hence, this paper aims to propose a CDSC method via parameter transferring and attention sharing mechanism (PTASM), and the presented architecture includes the source domain network (SDN) and the target domain network (TDN). First, hierarchical attentional network with pre-training language model on training data, such as global vectors for word representation and bidirectional encoder representations from transformers (BERT), are constructed. The word and sentence levels of parameter transferring mechanisms are introduced in the model transfer. Then, parameter transfer and fine-tuning techniques are adopted to transfer network parameters from SDN to TDN. Moreover, sentiment attention can serve as a bridge for sentiment transfer across different domains. Finally, word and sentence level attention mechanisms are introduced, and sentiment attention is shared from the two levels across domains. Extensive experiments show that the PTASM-BERT method achieves state-of-the-art results on Amazon review cross-domain datasets.

Prof. YADAV Devinder Kumar

Previously employed at the University of Nottingham/School of Aerospace, Australia

Biography: Devinder, an aeronautical engineer, has extensive experience in aviation industry and academia. He has held a variety of responsible leadership, management, and academic positions in large international airlines, aircraft manufacturers, and universities. He is an Australian national. In addition to the academic qualifications, he also holds an Aircraft Engineer License with several ratings starting from Cessna 152 to wide-body Airbus & Boeing 747-400 aircraft, and a Pilot License. Furthermore, he holds a range of other aviation professional qualifications in the field of aviation management, auditing, quality & safety, human factors, dangerous goods, training the trainer, and accident/incident investigation. His publications in scholarly international journals cover interdisciplinary areas, such as aviation regulations, flight safety, aircraft personnel licensing, airport engineering, higher education, airworthiness, aircraft maintenance, aviation management, and project management.

Title: A comparative study of touchscreen-based aircraft instruments and conventional instruments in relation to human factor.

Abstract: Integrated aircraft instruments are determinantal to flight safety and pilot’s workload. This lecture discusses a comparison between touchscreen-based aircraft instruments and conventional instruments used for navigation and flight profile monitoring in terms of pilot distraction and possible human errors. This study was carried out on laboratory-based flight simulators at a university facility. Several simulated flights using four pilots were operated to observe the human factor errors. Aircraft speed, altitude and heading parameters were selected for the observation, and the analysis was done using the Mean Squared Error (MSE) mathematical model. Videos of each flight were also made to visually observe the distraction. A relationship between flight task-time and parameters fluctuation as a result of human error was noticed. The errors were fewer using touchscreen instruments as compared to the conventional ones. The pilots' age was not considered for this study.

A. P. Xiang LI, Department of Automation, Tsinghua University

Biography: Xiang LI is an Associate Professor with the Department of Automation, Tsinghua University. He has been the Associate Editor of IEEE Robotics and Automation Letters since 2022 and the Associate Editor of IEEE Transactions on Automation Science and Engineering since 2023. He was the Associate Editor of IEEE Robotics & Automation Magazine from 2019 to 2021 and the Associate Editor of ICRA from 2019 to 2021. He received the Highly Commended Paper Award in 2013 IFToMM, the Best Paper in Robotic Control in 2017 ICAR, the Best Application Paper Finalists in 2017 IROS, the T. J. Tarn Best Paper in Robotics in 2018 IEEE ROBIO, and the Best Paper Award in 2023 ICRA DOM Workshop. His current research interests include robotic manipulation, vision-based control, micro/nano robots, and human-robot interaction.

Title: Robotic Manipulation of Deformable Linear Objects

Abstract: Robotic manipulation of deformable linear objects (DLOs) has broad application prospects in many fields, such as manufacturing and surgery. A key challenge is that it is difficult to obtain the exact deformation models of DLOs which describe how the robot motion affects the deformation. This is because the models are hard to theoretically calculate and vary among different DLOs. Thus, shape control of DLOs is challenging, especially for large deformation shape control, which requires global and more accurate models. This talk proposes a coupled offline and online data-driven method for efficiently learning a global deformation model, allowing for both more accurate modeling through offline learning and further updating for new DLOs via online adaptation. Real-world experiments and applications of the proposed method are presented.