Cloud-Edge-Device Collaborative Intelligent Model Technology Forum
August 22 (Friday) 3:30 PM-5:30 PM
Location: VIP Room, 3rd Floor

Introduction:
        Professor, a National Outstanding Young Scientist and a Thousand Talents Awardee, received his bachelor's and doctoral degrees from Tsinghua University and the Hong Kong University of Science and Technology. He has extensively researched distributed systems and intelligent computing centers, and the cloud computing and AI big-model applications they support. He has undertaken major national laboratory projects, national key R&D program projects, and projects funded by the National Natural Science Foundation of China. He led his team, leveraging China's proprietary and controllable computing power, to develop a 33 billion-parameter long-window big-model and a full-lifecycle toolset. These results have been applied to renowned companies such as Huawei, Baichuan, Inspur, ICBC, and TCL. He has won the second prize of the National Natural Science Award, the first prize of the Natural Science Award of the Ministry of Education, the Young Scientist Award of the Ho Ying-Tong Education Foundation, and the champion award of the "Code to Space" track in the national finals of the Huawei Developer Competition. He has published more than 160 high-level papers (including AAAI, WWW, SIGCOMM, SIGMETRICS, ISCA, ASPLOS, ICLR, ACL, USENIX FAST, ATC, MLSys, IEEE/ACM Transactions and other top conferences and journals), and was selected into the top 2% of the world's top scientists and Elsevier's highly cited scholars in China.

Report Title: Opportunities, Challenges, and Application Prospects of Domestic Computing Power and Large AI Models

Report Introduction:  

        Faced with the opportunities and challenges brought by the rise of large AI models and intelligent agents, the global intelligent computing and cloud computing industries and the technology community are vying for supremacy. This report explores how China's domestic cloud brain computing platform can support the formation of a full chain of large models and tool sets in China, with independent control, demonstration applications, and continuous evolution. It also shares the scientific and technological challenges discovered in the practice of training large models with a scale of hundreds of billions, and the progress of building organized innovation teams.

Introduction:
       Changjiang Special Post Scholar, Distinguished Professor at Shenzhen University. Main research area: general big data processing. He holds a bachelor's and master's degree in computer science from the University of Science and Technology of China, a master's degree in statistics, and a doctorate in computer science from the University of Texas at Austin. He was formerly a senior engineer at Oracle USA and is currently the Deputy Director of the National Engineering Laboratory for Big Data System Computing Technology and the Executive Director of the Shenzhen Institutes of Computational Science. He proposed a big data generalization model to address the challenges of diversity and established a general theoretical framework for big data management and analysis based on metric spaces. His innovative achievements have been recognized with the 2014 Ministry of Education Science and Technology Progress Award (Second Prize), the 2016 Military Science and Technology Progress Award (Second Prize), the 2021 Guangdong Province Teaching Achievement Award (First Prize), the 2022 National Teaching Achievement Award (Second Prize), the 2023 "National Huang Danian Teacher Team" award, the 2023 China Computer Society Natural Science Award (Second Prize), the 2023 China Electronics Society Natural Science Award (Second Prize), and the 2024 China Electronics Society Natural Science Award (First Prize), among other national, provincial, ministerial, and authoritative academic and industry awards.

Report Title: Universal representation of graph data based on metric space

Report Abstract:  

        One of the core research areas for pre-training models is the development of universal representations that can be used multiple times. Traditional machine learning is often limited to the Euclidean norm, which falls short of the inherent non-Euclidean nature of graphs. Metric spaces do not constrain the internal structure of the data; they only require that the distances between data satisfy positivity, symmetry, and trigonometric inequality, allowing them to represent a wide range of graph data. We propose a new paradigm that first represents the graph as a metric space and then vectorizes it, followed by training and fusing models under multiple norms. We investigate a theoretical framework for representation learning in metric spaces, including universal approximations under multiple norms, model parameter optimization mechanisms, and multimodal/multi-task training fusion mechanisms. This research is expected to explore new avenues for universal representation of graph data.

Introduction:
        Dr. Wu Fan is currently a Distinguished Professor and Executive Vice Dean of the School of Computer Science at Shanghai Jiao Tong University, and a recipient of the National Natural Science Foundation of China Outstanding Young Scientist Fund. He has published over 300 academic papers in the fields of mobile intelligent computing, large-scale model collaboration, and wireless networks. He has served on the editorial boards of five international academic journals, including IEEE Transactions on Mobile Computing. He has won two first-class Natural Science Awards from the Ministry of Education, a first-class Shanghai Science and Technology Progress Award, and seven international conference paper awards. He is the project leader for major projects such as the Science and Technology Innovation 2030 "New Generation Artificial Intelligence" project and key projects of the National Natural Science Foundation of China.

Report Title: Lightweight intelligent computing on mobile devices

Report Introduction:

Introduction:
      Professor and doctoral supervisor at Beijing University of Posts and Telecommunications. Recipient of the National Outstanding Young Scientist Fund and the 18th China Youth Science and Technology Award. Vice Chairman of the Space-Time Security Information Service Professional Committee of the Chinese Command and Control Society. He has presided over more than 20 projects, including the National Key R&D Program, the National Natural Science Foundation of China Enterprise Joint Fund, the National Defense Innovation Zone Project, and the National Defense Science and Technology 173 Program Key Project. He has won the IEEE ICC 2022 and other conference best paper awards 6 times, the Ministry of Education Technology Invention First Prize, the China Electronics Society Technology Invention First Prize, the China Invention Association Invention and Entrepreneurship Award Innovation First Prize, the Wu Wenjun Artificial Intelligence Science and Technology Progress Award Second Prize, etc. He serves as the deputy editor of IEEE Transactions on Sustainable Computing, the deputy editor of IEEE Transactions on Circuits and Systems for Video Technology, the guest editor of IEEE Open Journal of the Computer Society, and a member of the editorial board of "Integrated Space Information Network".

Report Title:

Report Introduction:

Introduction:
       Tenured Professor, Doctoral Supervisor, and Director of the Institute of High Performance Computing, Department of Computer Science, Tsinghua University. Dean of the School of Computer Technology and Applications, Qinghai University. Recipient of the National Outstanding Young Scientist Fund. Deputy Director of the CCF High Performance Computing Committee and Distinguished Member of CCF. His main research areas include parallel computing, programming models, and compilation optimization. He has published more than 100 papers in top conferences and journals in the field of parallel computing and systems, and published one monograph. His research results have won the IEEE TPDS 2021 Best Paper Award, the IEEE CLUSTER 2021 Best Paper Award, the ACM ICS 2021 Best Student Paper Award, etc. He served as the Chair of the NPC 2018 Program Committee, the Area Chair of IEEE CLUSTER 2021, and an editorial board member of several international academic journals such as IEEE Transactions on Computers. He serves as the coach of the Tsinghua University Student Supercomputing Team, and the team he coached has won the world championship fifteen times. He has won the First Prize of Science and Technology Progress of the Ministry of Education, the First Prize of Natural Sciences of the China Computer Society, the CCF-IEEE CS Young Scientist Award, the Outstanding Teacher Award Program for Computer Science in Colleges and Universities, and the Okawa Foundation of Japan.

Report Title: Large model reasoning system

Report Introduction:

        Large models empower countless industries, but their inference systems, as the supporting engines, face the challenge of high inference costs. This report will deeply analyze the key technologies of large-model inference systems from four key perspectives: memory management, compilation optimization, model quantization, and parallelization strategies. This report will explore efficient memory management methods, compilation optimization, model compression and quantization, and parallelization strategies to provide a reference for building efficient and low-cost large-model inference systems. At the same time, we have developed the "Chitu" large-model inference engine for domestic intelligent computing power, effectively improving the inference efficiency of large models on domestic computing power.