The 2023 Annual ShanghaiTech Symposium on Information Science and Technology
(ASSIST 2023)

The 2023 Annual ShanghaiTech Symposium on Information Science and Technology
(ASSIST 2023)

23-24 September 2023, Shanghai, China

23-24 September 2023, Shanghai, China


Program

Keynote Speakers

Weinan E

Peking University

Academician of Chinese Academy of Sciences
Chair Professor of Peking University
Co-Director, National Engineering Laboratory of Big Data Analysis and Applications
Director of Center for Machine Learning Research Peking University
Dean of AI for Science Institute. Beijing
Speech detail

Alessandro Foi

Faculty of Information Technology and Communication Sciences, Tampere University

Director of Tampere University Imaging Research Platform
Editor-in-Chief of IEEE Transactions on Image Processing
IEEE Fellow, for contributions to image restoration and noise modeling
Professor of Signal Processing
Speech detail

Boon Thau Loo

Electrical and Systems Engineering, University of Pennsylvania

Associate Dean, Graduate Programs – School of Engineering and Applied Science
Director, Distributed Systems Laboratory
Professor
Speech detail

Dacheng Tao

School of Computer Science, University of Sydney

Chief Scientist and Advisor of Institute of Digital Science
Fellow of the Australian Academy of Sciences
IEEE, ACM and AAAS Fellow
Professor
Speech detail

 

Invited Speakers

Xiao-Chuan Cai

Director of Centre for Applied Mathematics, University of Macau

Associate Dean facualty of Science and Technology
SIAM Fellow
Chair Professor
Speech detail

Yong (Peter) Lian

York University, Canada

Fellow of the Canadian Academy of Engineering
Fellow of Academy of Engineering, Singapore
IEEE Fellow
Professor
Speech detail

Pietro Lio

Department of Computer Science and Technology, Cambridge University

Member of the European Academy of Sciences
Member of the Artificial Intelligence group
Professor
Speech detail

Chen Change Loy

School of Computer Science and Engineering, Nanyang Technological University

Lab Director of MMLab@NTU
Co-Associate Director of S-Lab for Advanced Intelligence
Professor
Speech detail

 

Wayne Luk

Department of Computing, Imperial College London

Fellow of the Royal Academy of Engineering
IEEE Fellow
Professor
Speech detail

Mohamad Sawan

School of Life Science, Westlake University

Fellow of the Canadian Academy of Engineering
IEEE Fellow
Professor
Speech detail

Dinggang Shen

School of Biomedical Engineering, Shanghaitech University

Founding Dean, School of Biomedical Engineering
Director of IDEA Laboratory
IEEE and AIMBE Fellow
Professor
Speech detail

Martin Steinegger

Biology department, Seoul National University

Assistant Professor of Bioinformatics
Speech detail

 

Qi Tian

Huawei

Chief Scientist, Artificial Intelligence, Huawei Cloud
Member of the International Eurasian Academy of Sciences
IEEE Fellow
Speech detail

He Wang

School of Computer Science, Peking University

Assistant Professor at Center on Frontiers of Computing Studies
Founder and Leader of Embodied Perception and InteraCtion (EPIC) Lab at Peking University
Director of Embodied AI Center, Beijing Academy of Artificial Intelligence (BAAI)
Professor
Speech detail

Yu Wang

Department of Electronic Engineering, Tsinghua University

Chair of Department of Electronic Engineering
IEEE Fellow
Speech detail

Qi Wu

School of Computer and Mathematical Sciences, University of Adelaide

ARC DECRA Fellow
Associate Professor
Speech detail

 

Jinbo Xu

Institute for AI Industry Research, Tsinghua University

Editorial board member of Bioinformatics
Professor

Jianyi Yang

Research Center for Mathematics and Interdisciplinary Science, Shandong University

Professor
Speech detail

Yang You

Department of Computing Science, National University of Singapore

Presidential Young Professor
Speech detail

Xiaoyang Zeng

School of Information Science and Technology, Fudan University

Vice Dean of School of Microelectronics
Professor
Speech detail

 

Fumin Zhang

Hong Kong University of Science and Technology

Director of HKUST Cheng Kar-Shun Robotics Institute
Chair Professor
Speech detail

Hanwang Zhang

School of Computer Science and Engineering, Nanyang Technological University

Associate Professor
Speech detail

Yue Zhang

Westlake University

Professor
Speech detail

Bolei Zhou

Computer Science Department, University of California, Los Angeles

Assistant Professor
Speech detail

 

Jun Zhou

School of Information and Communication Engineering, University of Electronic Science and Technology of China

National-level young talent
Professor
Speech detail

S Kevin Zhou

University of Science and Technology of China

IEEE and AIMBE Fellow
Chair Professor
Speech detail

Feng Zhu

College of Pharmaceutical Sciences, Zhejiang University

Editor-in-Chief, Computers in Biology and Medicine
Professor
Speech detail

Jun Zhu

Computer Science Department, Tsinghua University

Vice President of Institute for Artificial Intelligence, Tsinghua University
Associate Editor for Artificial Intelligence.
IEEE Fellow
Professor
Speech detail

 

Chengqing Zong

Institute of Automation, Chinese Academy of Sciences

IEEE and ACL Fellow
Professor
Speech detail

 

 

 

 

Speakers and Speeches Information

Xiao-Chuan Cai

Director of Centre for Applied Mathematics, University of Macau

Title: Computational Technologies for Understanding the Human Cardiovascular System

Abstract:  We present a highly parallel domain decomposition algorithm for the simulation of blood flows in the human body governed by the incompressible Navier-Stokes equations. The system is discretized with a fully implicit finite element method on unstructured moving meshes in 3D and solved by a Newton-Krylov algorithm preconditioned with an overlapping Schwarz method on large scale supercomputers. Several mathematical, bio-mechanical, and supercomputing issues will be discussed in detail, and some numerical experiments for the simulation of the human cardiovascular system will be presented.

Bio:  X.-C. Cai's research interest is large scale scientific computing and computational biomechanics. He received his BSc in 1984 from Peking University, and PhD in 1989 from Courant Institute of Mathematical Sciences, New York University. He was a postdoc during 1989-1990 at Yale University. He is currently a Chair Professor in the Department of Mathematics, University of Macau. He is a fellow of SIAM.

Weinan E

Peking University

Title: AI for Science

Abstract:  For many years, the lack of good algorithms has severely limited our ability to conduct scientific research. At its heart, the difficulty comes from the notorious “curse of dimensionality” problem. Deep learning is exactly the kind of tool needed to address this problem. In the last few years, we have seen a tremendous amount of scientific progress made as a result of the AI revolution, both in our ability to make use of the fundamental principles of physics, and our ability to make use of experimental data. In this talk, I will start with the origin of the AI for Science revolution, review some of the major progresses made so far, and discuss how it will impact the way we do scientific research. I will also discuss how AI for Science might impact applied mathematics.

Bio:  Weinan E is a professor in the Center for Machine Learning Research (CMLR) and the School of Mathematical Sciences at Peking University. He is also the inaugural director of the AI for Science Institute in Beijing, as well as the director of the Beijing Institute for Big Data Research. He is a member of the Chinese Academy of Sciences; a fellow of SIAM, AMS, IOP, CSIAM and ORSC. His main research interest is numerical algorithms, machine learning and multi-scale modeling, with applications to chemistry, material sciences and fluid mechanics. He was a plenary speaker at the 2022 International Congress of Mathematicians (ICM), a keynote speaker at the 2022 International Conference on Machine Learning (ICML) and an invited speaker at ICM2002 and ICIAM (International Congress of Industrial and Applied Mathematics) 2007. He has also been invited speaker at leading conferences in many other scientific disciplines, including the APS, ACS, AIChe annual meetings, the American Conference of Theoretical Chemistry and the World Congress of Computational Mechanics. He was awarded the ICIAM Collatz Prize in 2003, the SIAM Kleinman Prize in 2009, the SIAM von Karman Prize in 2014, the SIAM-ETH Peter Henrici Prize in 2019, the ACM Gordon-Bell Prize in 2020, and the ICIAM Maxwell Prize in 2023.

Alessandro Foi

Faculty of Information Technology and Communication Sciences, Tampere University

Title: A platform to foster and develop collaborative cross-disciplinary and cross-sector research and innovation in imaging

Abstract:  The impact and societal integration of academic research organizations is often limited by the adoption of simplistic performance indicators tied to bibliometry. This talk presents the Imaging Research Platform established in Finland at Tampere University as an operational model to foster and develop cross-disciplinary and cross-sector research and innovation, with the longer-term ambition of making impact beyond academic publishing. We touch upon aspects such as the Platform's operational structure, the design of internal seed-funding calls, formation of cross-disciplinary teams, projects deliverables, mentoring, follow-up, and measures of impact. On a broader scale, we discuss the vertical integration into the degree education and the industry innovation ecosystem, the support of research infrastructures, the interaction with government and public stakeholders, outreach, and the international and cross-sector mobility of researchers. The talk will also present some of the Platform's research highlights from the 'Advanced Imaging as a Service' model and from the research collaboration on 'Light Transport'.

Bio:  Alessandro Foi received the M.Sc. degree in Mathematics from the Università degli Studi di Milano, Italy, in 2001, the Ph.D. degree in Mathematics from the Politecnico di Milano in 2005, and the D.Sc.Tech. degree in Signal Processing from Tampere University of Technology, Finland, in 2007. He is a Professor of Signal Processing at Tampere University, Finland, and the Director of Tampere University Imaging Research Platform. His research interests include mathematical and statistical methods for signal processing, functional and harmonic analysis, and computational modeling of the human visual system. His work focuses on spatially adaptive (anisotropic, nonlocal) algorithms for the restoration and enhancement of digital images, on noise modeling for imaging devices, and on the optimal design of statistical transformations for the stabilization, normalization, and analysis of random data. He is the Editor-in-Chief of the IEEE Transactions on Image Processing. He previously served as a Senior Area Editor for the IEEE Transactions on Computational Imaging and as an Associate Editor for the IEEE Transactions on Image Processing, the SIAM Journal on Imaging Sciences, and the IEEE Transactions on Computational Imaging. He is a Fellow of the IEEE.

Yong (Peter) Lian

York University, Canada

Title: Event-Driven Circuits and Systems for Ultra Low Power Artificial Intelligent IoT Chips

Abstract:  Artificial Intelligence-of-Things (AIoT) devices differ from the IoT counterpart that not only they sense, store, transmit data but also analyze and act on data, i.e. they perform tasks similar to what a person could do. For the ubiquitous deployment of sensors in AIoT, power consumption of each sensor should be made as low as possible for extended operation hours. The traditional signal processing flow relies on Nyquist rate to digitize input signal, which generates redundant samples for sparse inputs in many AIoT applications and increases power consumption for the AI engine. Different from Nyquist sampling scheme, the event-driven system generates samples only if a predefined event occurs, i.e. the power consumption tracks the input activities. The amount of data generated in an event-driven system is intrinsically compressed, leading to significant savings in power. This talk will introduce several low power features of an event-driven system, and demonstrate how to achieve orders of magnitude reduction in power consumption for AIoT chips.

Bio:  Peter Lian is a Fellow of the Canadian Academy of Engineering, the Academy of Engineering Singapore, and IEEE. His research interests are in the areas of self-powered intelligent wireless sensors, energy efficient event-driven signal processing techniques and biomedical circuits and systems. He is the recipient of more than 15 awards including the 2023 IEEE Circuits and Systems Society Mac Van Valkenburg Award, 2023 IEEE Transactions on Biomedical Circuits and Systems Best Paper Award, Design Contest Award at 2015 International Symposium on Low Power Electronics and Design, 2011 Institution of Engineers Singapore Prestigious Engineering Achievement Award, 2008 IEEE Communications Society Multimedia Communications Best Paper Award, and 1996 IEEE Circuits and Systems Society Guillemin-Cauer Award. He serves as the IEEE Division 1 (Microelectronics) Director-Elect, IEEE Board of Director-Elect, Member-at-Large of the IEEE Publication Services and Products Board, Chair for the IEEE Periodicals Partnership Opportunities Committee, Chair of the IEEE Ad Hoc Committee on Accelerating Partnerships with Chinese Publications, member of the IEEE Periodicals Committee, member of the IEEE MGA Strategic Planning Committee, member of the IEEE Periodicals Review and Advisory Committee, member of the IEEE PSPB Publishing Conduct Committee, and member of the IEEE Fellow Committee. He was the President of the IEEE Circuits and Systems Society, Editor-in-Chief of the IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II: EXPRESS BRIEFS, the VP for Publications and VP for Region 10 of the IEEE CAS Society, and many other roles in the IEEE.

Pietro Lio

Department of Computer Science and Technology, Cambridge University

Title: Actionable and responsible AI in medicine

Abstract:  In this talk I will focus on how to build a digital patient twin using graph and hypergraph representation learning and considering multiscale physiological (cardiovascular), clinical (inflammation) and molecular variables (multi omics and genetics). I will discuss methodologies that help keeping the clinicians in the loop to avoid excessive automatisation using logic and explainer frameworks.

Bio:  Pietro Lio received the PhD degree in complex systems and nonlinear dynamics from the School of Informatics, dept of Engineering, University of Firenze, Italy and the PhD degree in theoretical genetics from the University of Pavia, Italy. He is currently a professor of computational biology with the Department of Computer Science and Technology, University of Cambridge and a member of the Artificial Intelligence Group. He is also a member of the Cambridge Centre for AI in medicine, ELLIS (European Laboratory for Learning and Intelligent Systems), Academia Europaea, Asia Pacific Artificial Intelligence Association. His research interests include graph representation learning, AI and Medicine, Systems Biology.

Boon Thau Loo

Electrical and Systems Engineering, University of Pennsylvania

Title: Towards Full-Stack Adaptivity in Permissioned Blockchain Systems

Abstract:  Permissioned blockchain systems are an emerging instance of untrustworthy distributed databases. As novel smart contracts, modern hardware, and new cloud platforms arise, future-proof permissioned blockchain systems need to be designed with full-stack adaptivity in mind.  At the application level, a future-proof system must adaptively learn the best transaction processing paradigm in order to maximize performance for dynamic workloads, and quickly adapt to new hardware as well as unanticipated workload changes on-the-fly. Likewise, the Byzantine consensus layer must dynamically adjust itself to the workloads, faulty conditions, and network configuration while maintaining compatibility with the transaction processing paradigm. At the infrastructure level, cloud providers must enable cross-layer adaptation, which identifies performance bottlenecks and possible attacks, and determines at runtime the degree of resource disaggregation that best meets application requirements. This talk presents four preliminary building blocks towards our vision of full-stack adaptivity: (1) FlexChain, a novel permissioned blockchain system that physically disaggregating CPUs, DRAM, and storage devices to process different blockchain workloads efficiently; (2) AdaChain, a learning-based framework that adaptively chooses the best permissioned blockchain architecture to optimize effective throughput for dynamic transaction workloads; (3) Bedrock, a unified platform for Byzantine consensus protocol analysis, implementation, and experimentation; and (4) DeCon, a declarative programming language for implementing, optimizing, and verifying smart contracts deployed on Blockchain systems. We conclude the talk with our ongoing work towards the goal of full-stack adaptivity across transaction processing, consensus protocols, and hardware infrastructure layers. 

Bio:  Boon Thau Loo is the RCA Professor in the Computer and Information Science (CIS) department at the University of Pennsylvania. He is also the Associate Dean for Graduate Programs, where he oversees all academic and admissions operations for doctoral, master’s and professional programs at Penn Engineering. He received his Ph.D. degree in Computer Science from the University of California at Berkeley in 2006. Prior to his Ph.D., he received his M.S. degree from Stanford University in 2000, and his B.S. degree with highest honors from University of California-Berkeley in 1999. His research focuses on distributed data management systems, Internet-scale query processing, and the application of data-centric techniques and formal methods to the design, analysis and implementation of networked systems. He is the recipient of the David J. Sakrison Memorial Prize (2006) for the most outstanding dissertation research in the Department of EECS at University of California-Berkeley, and the ACM SIGMOD Dissertation Award (2007), NSF CAREER award (2009), the Air Force Office of Scientific Research (AFOSR) Young Investigator Award (2012), Penn’s Emerging Inventor of the year award (2018), the Ruth and Joel Spira award for Excellence in Teaching (2021), and the University Lindback award for distinguished teaching (2022). He has published 160+ peer reviewed publications and has supervised sixteen Ph.D. dissertations and three postdocs. His graduated doctoral students and postdocs include three tenured professors, four current tenure-track professors, and winners of five dissertation awards. As an entrepreneur, he co-founded two companies: Netsil, a cloud microservices analytics company acquired by public cloud company Nutanix Inc., and Termaxia, an energy-efficient big data storage company acquired by Frontiir.  

Chen Change Loy

School of Computer Science and Engineering, Nanyang Technological University

Title: Harnessing Diffusion Prior for Content Enhancement and Creation

Abstract:  This talk delves into the exploration and application of pretrained diffusion models for content enhancement and creation. By leveraging the abundant image priors and robust generative capability of diffusion models, we innovatively address diverse applications including face restoration, image super-resolution, image colorization, and video-to-video translation. Our work provides a novel approach to content enrichment by harnessing the inherent structure of visual data through the diffusion process. This strategy elucidates the unique potential of utilizing existing models in diverse domains without explicit retraining, thereby reducing computational overheads and enabling efficient adaptability. Through this discussion, we aim to provide insights into the viability of diffusion models as a powerful tool for image and video enhancement tasks, and stimulate further research in exploiting the generative potential of diffusion models.

Bio:  Chen Change Loy is a Nanyang Associate Professor with the School of Computer Science and Engineering, Nanyang Technological University (NTU), Singapore. He serves as the Co-Associate Director for S-Lab, NTU and Director of MMLab@NTU. He is also an Adjunct Associate Professor at the Chinese University of Hong Kong. He received his PhD (2010) in Computer Science from the Queen Mary University of London. Before joining NTU, he served as a Research Assistant Professor at the MMLab of the Chinese University of Hong Kong, from 2013 to 2018. He is the recipient of 2019 Nanyang Associate Professorship (Early Career Award) from Nanyang Technological University. He is recognized as one of the 100 most influential scholars in computer vision by AMiner. His research interests include computer vision and deep learning with a focus on image/video restoration, enhancement, and content creation. He serves as an Associate Editor of the IJCV and TPAMI. He also serves/served as the Area Chair of top conferences such as CVPR, ICCV, and ECCV. He is a senior member of IEEE.

Wayne Luk

Department of Computing, Imperial College London

Title: Hardware Acceleration for Machine Learning Applications

Abstract:  This talk describes recent advances in hardware acceleration for a variety of high-performance applications involving machine learning. Such applications include adaptive radiotherapy, high energy physics, and financial risk management. Parametric architectural templates and the associated optimisations have been developed for various machine learning methods, such as convolutional neural networks, recurrent neural networks, graph neural networks, spiking neural networks, transformers, reinforcement learning, Gaussian mixture model, and inductive logic programming. Related techniques, including those for supporting uncertainty estimation, design space exploration and diverse implementation generation, will also be presented.

Bio:  Wayne Luk is Professor of Computer Engineering at Imperial College. He was a Visiting Professor at Stanford University. His research focuses on theory and practice of customizing hardware and software designs for demanding applications. He is a Fellow of the Royal Academy of Engineering, the IEEE, and the BCS.

Mohamad Sawan

School of Life Science, Westlake University

Title: Closed-loop neuromodulation-based Medical Devices to Diagnosis, Treat and Predict Brain Disorders

Abstract:  Closed-Loop neuromodulation based medical devices  intended for efficient diagnosis and treatment of neurodegenerative diseases are targets to mimic brain regular operation. Consequently, artificial intelligence-based learning techniques are the heart parts of these emerging control units to be embedded in proposed neuromodulation systems. This talk covers the implementation of wearable and implantable medical devices based on custom system-on-chip (SoC) integrated platforms.  The latter are intended for the diagnosis, treatment, and prediction of health conditions. These devices include signal processing methods, design and tests of SoCs and system assembly of bioelectronic closed-loop systems for brain interfaces. These methods deal with multidimensional design challenges such as efficient power management, very low-power and high-data rate wireless communication methods, and reliable systems. In these neuromodulation applications, priority could be given to non-invasive approaches, however for some healthcare dysfunctions, wearable systems can not apply, implantable devices should be used. Also, optoelectronic methods are used to build proposed closed-loop systems for both non-invasive nanoimaging, and transcranial stimulation. Case studies include several applications such as epilepsy, vision, addictions, and early and fast viruses detection.

Bio:  Mohamad Sawanis Chair Professor in Westlake University, Hangzhou, China, and Emeritus Professor in Polytechnique Montreal, Canada. He is founder and director of the Center of Excellence in Biomedical Research on Advanced Integrated-on-chips Neurotechnologies (CenBRAIN Neurotech) in Westlake University, Hangzhou, China. Also, he is founder of the Polystim Neurotech Laboratory in Polytechnique Montréal. He received the Ph.D. degree from University of Sherbrooke, Canada. Dr. Sawan research activities are bridging micro/nano electronics with biomedical engineering to introduce smart medical devices dedicated to improving the quality of human life. He is co-founder and was Editor-in-Chief of the IEEE Transactions on Biomedical Circuits and Systems (2016-2019). He hosted the 2016 IEEE International Symposium on Circuits and Systems, and the 2020 IEEE International Medicine, Biology and Engineering Conference (EMBC). He was a Canada Research Chair in Smart Medical Devices (2001-2015), and was leading the Microsystems Strategic Alliance of Quebec, Canada (1999-2018). Dr. Sawan published more than 1000 peer reviewed papers and many books and patents. Among the numerous received honors, Dr. Sawan received the Chinese National Friendship Award, The Lebanese’s President Medal of Merit, the Shanghai International Collaboration Award, the Queen Elizabeth II Golden Jubilee Medal. Dr. Sawan is Fellow of the Royal Society of Canada, Fellow of the Canadian Academy of Engineering, Fellow of the IEEE, and “Officer” of the National Order of Quebec.

Dinggang Shen

School of Biomedical Engineering, Shanghaitech University

Title: Fast Development and Deployment of AI Techniques for Medical Imaging

Abstract:  I will introduce our developed full-stack, full-spectrum Artificial Intelligence (AI, or deep learning) techniques for whole clinical workflow, from data acquisition to disease detection, follow-up, diagnosis, therapy, and outcome prediction (or evaluation). In particular, I will demonstrate some innovative technical development and implementation in scanners and clinical pipelines, i.e., serving for fast MR, low-dose CT/PET acquisition, and clinical diagnosis/therapy.

Bio:  Dinggang Shen is a Professor and a Founding Dean with School of Biomedical Engineering, ShanghaiTech University, Shanghai, China, and also a Co-CEO of United Imaging Intelligence (UII), Shanghai. He is a Fellow of IEEE, AIMBE, IAPR, and MICCAI. He was also a recipient of the Distinguished Investigator Award from The Academy for Radiological & Biomedical Imaging Research, USA (2019). He was Jeffrey Houpt Distinguished Investigator and a Full Professor (Tenured) with The University of North Carolina at Chapel Hill (UNC-CH), Chapel Hill, NC, USA, directing The Center of Image Analysis and Informatics, The Image Display, Enhancement, and Analysis (IDEA) Lab, and The Medical Image Analysis Core. He was also a tenure-track assistant professor in the University of Pennsylvanian (UPenn), and an Instructor in the Johns Hopkins University. His research interests include medical image analysis, computer vision, and pattern recognition. He has published more than 1600 peer-reviewed papers in the international journals and conference proceedings, with H-index 133 and over 75K citations. He serves as an Editor-in-Chief for Frontiers in Radiology, as well as an editorial board member for eight international journals. Also, he has served in the Board of Directors for MICCAI Society in 2012-2015, and was General Chair for MICCAI 2019.

Martin Steinegger

Biology department, Seoul National University

Title: Clustering predicted structures at the scale of the known protein universe

Abstract:  Sequence-based predictions of protein structures have increased in accuracy with over 214 million predicted structures available in the AlphaFold database (AFDB). However, studying protein structures at this scale requires highly efficient methods. Here, we developed a structural-alignment based clustering algorithm – Foldseek cluster - that can cluster hundreds of millions of structures. Using this method we have clustered all structures in AFDB, identifying 2.27M non-singleton structural clusters, of which 31% lack annotations representing likely novel structures. Clusters without annotation tend to have few representatives covering only 4% of all proteins in the AFDB. Evolutionary analysis suggests that most clusters are ancient in origin but 4% seem species specific, representing lower quality predictions or examples of de-novo gene birth. Additionally, we show how structural comparisons can be used to predict domain families and their relationships, identifying examples of remote homology. Based on these analyses we identify several examples of human immune related proteins with remote homology in prokaryotic species which illustrates the value of this resource for studying protein function and evolution across the tree of life. Easy exploration of the clusters is available at: https://cluster.foldseek.com

Bio:  Dr. Steinegger is an Assistant Professor at the Seoul National University, where he is affiliated with the Biology department, Institute of Molecular Biology and Genetics, Artificial Intelligence Institute, and the Bioinformatics Graduate School. His research group focuses on developing big data and machine learning algorithms to analyze genomic and proteomic sequence data. The group is best known for bioinformatics software to cluster (Linclust), assemble (Plass), search (MMseqs2) sequences and to predict protein structures (AlphaFold2/ColabFold), search (Foldseek), and compress (Foldcomp) them. These software packages are used by researchers worldwide and have been installed hundreds of thousands of times. Dr. Steinegger is an expert on large-scale sequence data analysis and method development and advocates for open science and open source.

Dacheng Tao

School of Computer Science, University of Sydney

Title: More Is Different: Beyond Wittgenstein’s Philosophy

Abstract:  Delving into the hidden wisdom within vast data is an enticing pursuit. The key: foundation models. These powerful transformers unlock the enigmatic knowledge concealed in data’s depths. They combine parameters, computations, and data in a symphony of potential, reigniting our quest for Artificial General Intelligence. In this presentation, we embark on a thrilling journey into the realm of foundation models. Starting with the ground-breaking LLMs, such as ChatGPT, we explore the innovation, trend, and opportunity they’ve ignited. Along the way, we grapple with concerns surrounding singularity while offering unique insights into this burgeoning trend. We delve into the theoretical bedrock, explore ingenious decentralized optimization algorithms, and unearth the fertile ground where applications flourish under the influence of these models. This adventure doesn’t shy away from the challenges and opportunities that await in the era of these titans. Our optimism remains steadfast: foundation models are poised to be the architects of the of AI within the next five years. Let’s rock on this remarkable expedition where data, computational prowess, and algorithms converge to reveal unparalleled possibilities.

Bio:  Dacheng Tao is currently a Professor of Computer Science, Peter Nicol Russell Chair and an Australian Laureate Fellow in the Sydney AI Centre and the School of Computer Science in the Faculty of Engineering at The University of Sydney. He was the founding director of the Sydney AI Centre. He mainly applies statistics and mathematics to artificial intelligence and data science, and his research is detailed in one monograph and over 200 publications in prestigious journals and proceedings at leading conferences. He received the 2015 and 2020 Australian Eureka Prize, the 2018 IEEE ICDM Research Contributions Award, and the 2021 IEEE Computer Society McCluskey Technical Achievement Award. He is a Fellow of the Australian Academy of Science, AAAS, ACM and IEEE.

Qi Tian

Huawei

Title: Pangu Weather: Accurate Medium-range Global Weather Forecasting with 3D Neural Networks

Abstract:  Weather forecasting is important for science and society. Currently, the most accurate forecast system is the numerical weather prediction (NWP) method, which represents atmospheric states as discretized grids and numerically solves partial differential equations (PDEs) that describe the transition between those states. However, this procedure is computationally expensive. Recently, AI-based weather forecasting methods have shown potential in accelerating weather forecasting by orders of magnitudes, but the forecast accuracy is still significantly lower than that of NWP methods. In this paper, we introduce an AI-based method for accurate, medium-range global weather forecasting. We show that 3D deep networks equipped with Earth-specific priors are effective at dealing with complex patterns in weather data, and that a hierarchical temporal aggregation strategy reduces accumulation errors in medium-range forecasting. Trained on 39 years of global data, our program, Pangu-Weather, is the first to obtain stronger deterministic forecast results on reanalysis data in all tested variables, when compared with the world’s best NWP system, the operational integrated forecasting system (IFS) of the European Centre for Medium-Range Weather Forecasts (ECMWF). Our method also works well with extreme weather forecasts and ensemble forecasts. When initialized with reanalysis data, the accuracy of tracking tropical cyclones is higher than ECMWF-HRES.

Bio: 

He Wang

School of Computer Science, Peking University

Title: Towards 3D-Aware Multi-Function Home Robots

Abstract:  Home robots, as demonstrated by RT-1 and PaLM-E, have recently drawn much attention. Taking 2D images and language queries as input and directly outputting actions, these models are learned in a supervised manner, leading to two critical bottlenecks. First, they are not scalable. These models can handle only very limited tasks (mainly using parallel grippers to pick and place) but the training already requires a tremendous amount of expert demonstrations manually collected in the real world. Second, they lack of 3D vision and thus their interaction skills are unsatisfactory and limited in many ways. To build a general-purpose home robot in a scalable and extendable way, we thus prefer a modular-based design that builds upon several critical 3D-aware building blocks, i.e. navigation, object grasping, and object manipulation, rather than an end-to-end model. In this talk, I will focus on our recent works that significantly improve these modules and scalably extend their applicability by leveraging 3D vision techniques and synthetic data. Our work 3D-Nav proposes the first object goal navigation policy that utilizes online 3D reconstruction and scene understanding. Utilizing domain randomized synthetic data, our DREDS and GraspNeRF tackle the challenging real-world transparent object grasping problems and propose a depth restoration and a no-depth solution, respectively. Finally, for object manipulation, our CVPR’23 highlight work, GAPartNet, proposes generalizable and actionable parts, a.k.a GAPart, that facilitate cross-category object perception and manipulation.

Bio:  Dr. He Wang is a tenure-track assistant professor in the Center on Frontiers of Computing Studies (CFCS) at Peking University, where he founds and leads Embodied Perception and InteraCtion (EPIC) Lab. His research interests span 3D vision, robotics, and machine learning, with a special focus on embodied AI. His research objective is to endow robots working in complex real-world scenes with generalizable 3D vision and interaction policies in a scalable way. He has published more than 40 papers in top conferences and journals of computer vision, robotics, and learning, including CVPR/ICCV/ECCV/TRO/ICRA/IROS/NeurIPS/ICLR/AAAI. His pioneering work on category-level 6D pose estimation, NOCS, receives the 2022 World Artificial Intelligence Conference Youth Outstanding Paper (WAICYOP) Award and his work also receives ICRA 2023 outstanding manipulation paper award finalist and Eurographics 2019 best paper honorable mention. He serves as an associate editor of Image and Vision Computing and serves as an area chair in CVPR 2022 and WACV 2022. Prior to joining Peking University, he received his Ph.D. degree from Stanford University in 2021 under the advisory of Prof. Leonidas J. Guibas and his Bachelor's degree from Tsinghua University in 2014.

Yu Wang

Department of Electronic Engineering, Tsinghua University

Title: Sparse Acceleration for DNNs and LLMs: Progress and Trends

Abstract:  After decades of advancements, artificial intelligence algorithms have become increasingly sophisticated, with sparse computing playing a pivotal role in their evolution. This talk will first review and summarize the characteristics of sparsity in AI 1.0 & 2.0. Subsequently, three representative AI algorithms – Convolutional Neural Network, Graph Neural Network, and Large Language Model, along with their sparse computing optimizations, will be discussed. By analyzing the similarities and differences of these AI algorithms, we will propose the development trend of sparse computing. Finally, this talk will envision a future, where a coordinated dataflow, instruction-set-architecture, and hardware co-design approach would provide a great opportunity for efficient and general processing of various sparsity and sparse AI algorithms.

Bio:  Yu Wang, professor, IEEE fellow, chair of the Department of Electronic Engineering of Tsinghua University, dean of Institute for Electronics and Information Technology in Tianjin, and vice dean of School of information science and technology of Tsinghua University. His research interests include the application specific heterogeneous computing, processing-in-memory, intelligent multi-agent system, and power/reliability aware system design methodology. Yu Wang has published more than 80 journals (60 IEEE/ACM journals) and 200 conference papers in the areas of EDA, FPGA, VLSI Design, and Embedded Systems, with the Google citation more than 17,400. He has received four best paper awards and 12 best paper nominations. Yu Wang has been an active volunteer in the design automation, VLSI, and FPGA conferences. He will serve as TPC chair for ASP-DAC 2025. He serves as the editor of important journals in the field such as ACM TODAES and IEEE TCAD and program committee member for leading conferences in the top EDA and FPGA conferences.

Qi Wu

School of Computer and Mathematical Sciences, University of Adelaide

Title: Human-Computer Conversational Vision-and-Language Navigation

Abstract:  The dynamic realm of Vision-and-Language Navigation (VLN) has garnered significant multidisciplinary interest, resonating within the domains of computer vision, natural language processing, and robotics. This presentation embarks on a comprehensive exploration of the VLN trajectory, tracing its inception to seminal benchmarks such as Room-to-Room (R2R). A pivotal catalyst within this evolution is the advent of Large Language Models (LLMs), exemplified by the transformative GPT-4. These LLMs have not only facilitated more natural and fluid human-machine interactions but also unlocked novel pathways for leveraging human-like language in guiding robots through intricate navigational tasks. The discourse commences by establishing the contextual framework of human-machine conversational dynamics, contextualizing the paradigm shift and its reverberations. Subsequently, a detailed exposition of our recent undertakings in the VLN domain is presented. This involves harnessing the prowess of LLMs to decode complex navigational instructions embedded within natural language, thereby elevating robotic navigational capabilities. The presentation serves as an illuminating window into the transformative potential of merging vision, language, and robotics.

Bio:  Dr Qi Wu is an Associate Professor at the University of Adelaide and was the ARC Discovery Early Career Researcher Award (DECRA) Fellow between 2019-2021. He is the Director of Vision-and-Language at the Australia Institute of Machine Learning. Australian Academy of Science awarded him a J G Russell Award in 2019. He obtained his PhD degree in 2015 and MSc degree in 2011, in Computer Science from the University of Bath, United Kingdom. His research interests are mainly in computer vision and machine learning. Currently, he is working on the vision-language problem, and he is primarily an expert in image captioning and visual question answering (VQA). He has published more than 100 papers in prestigious conferences and journals, such as TPAMI, CVPR, ICCV, ECCV. He is also the Area Chair for CVPR and ICCV.

Jianyi Yang

Research Center for Mathematics and Interdisciplinary Science, Shandong University

Title: Deep learning protein structure

Abstract:  Significant breakthroughs have been achieved in protein structure prediction, due to the development of deep learning algorithms, accumulation of big protein sequence/structure data, and advance of hardware. My research group has developed a few widely used deep learning-based protein structure prediction algorithms such as trRosetta and trRosettaX. Based on these advances, our group (Yang-Server) won the tertiary structure prediction in the 15th Critical Assessment of protein Structure Prediction (CASP15). I will briefly summarize the recent progress and future challenges.

Bio:  Jianyi Yang is a Professor in Shandong University. He has received a few prestigious rewards including the National Science Fund for Distinguished Young Scholars. His research interests include protein and RNA structure prediction. He has made significant contributions to the development of many widely used tools, such as trRosetta and I-TASSER. He has published over 60 papers in journals such as Nature Methods, PNAS, with >14000 citations.

Yang You

Department of Computing Science, National University of Singapore

Title: Colossal-AI: A Unified Deep Learning System For Large-Scale Parallel Training

Abstract:  The Transformer architecture has improved the performance of deep learning models in domains such as Computer Vision and Natural Language Processing. Together with better performance come larger model sizes. This imposes challenges to the memory wall of the current accelerator hardware such as GPU. It is never ideal to train large models such as Vision Transformer, BERT, and GPT on a single GPU or a single machine. There is an urgent demand to train models in a distributed environment. However, distributed training, especially model parallelism, often requires domain expertise in computer systems and architecture. It remains a challenge for AI researchers to implement complex distributed training solutions for their models. To solve this problem, we introduce Colossal-AI, which is a unified parallel training system designed to seamlessly integrate different paradigms of parallelization techniques including data parallelism, pipeline parallelism, multiple tensor parallelism, and sequence parallelism. Colossal-AI aims to support the AI community to write distributed models in the same way as how they write models normally. This allows them to focus on developing the model architecture and separates the concerns of distributed training from the development process. Colossal-AI is able to achieve 2x speedup over state-of-the-art distributed systems for GPT model training. The source code can be found at this https://github.com/hpcaitech/ColossalAI

Bio:  Yang You is a Presidential Young Professor at the National University of Singapore. He is on an early career track at NUS for exceptional young academic talents with great potential to excel. He received his PhD in Computer Science from UC Berkeley. His advisor is Prof. James Demmel, who was the former chair of the Computer Science Division and EECS Department. Yang You's research interests include Parallel/Distributed Algorithms, High Performance Computing, and Machine Learning. The focus of his current research is scaling up deep neural networks training on distributed systems or supercomputers. In 2017, his team broke the world record of ImageNet training speed, which was covered by the technology media like NSF, ScienceDaily, Science NewsLine, and i-programmer. In 2019, his team broke the world record of BERT training speed. The BERT training techniques have been used by many tech giants like Google, Microsoft, and NVIDIA. Yang You’s LARS and LAMB optimizers are available in industry benchmark MLPerf. He is a winner of IPDPS 2015 Best Paper Award (0.8%), ICPP 2018 Best Paper Award (0.3%), AAAI 2023 Distinguished Paper Award (0.14%), and ACM/IEEE George Michael HPC Fellowship. Yang You is a Siebel Scholar and a winner of Lotfi A. Zadeh Prize. Yang You was nominated by UC Berkeley for ACM Doctoral Dissertation Award (2 out of 81 Berkeley EECS PhD students graduated in 2020). He also made Forbes 30 Under 30 Asia list (2021) and won IEEE CS TCHPC Early Career Researchers Award for Excellence in High Performance Computing. For more information, please check his lab’s homepage at https://ai.comp.nus.edu.sg/

Xiaoyang Zeng

School of Information Science and Technology, Fudan University

Title: The multi-chiplet-module (MCM) CIM AI Processor and its Applications

Abstract:  However, with the slowdown of Moore’s Law, traditional computing chips are difficult to meet the performance and efficiency requirements of the emerging neural network algorithms. The Compute-in-Memory (CIM) technology breaks down the barriers between storage and computing units, and is expected to alleviate the “memory all” bottleneck. At present, CIM-based monolithic AI processor chips still face the challenge of performance scalability. In this speech, the speaker will introduce a multi-chiplet-module (MCM) CIM AI processor chip, which features the Computing-on-Memory-Boundary architecture, which completely eliminates on-chip weight movement; and the MCM system achieves post-fabrication performance scalability by integrating different numbers of chiplets.

Bio:  Prof. Xiaoyang Zeng received the Ph.D. (Hons.) degrees from Chinese Academy of Sciences in 2001. Then joined Fudan University as Post-Doctor researcher from 2001 to 2003. Prof. Zeng has been with Fudan University as a faculty since 2003, where he is currently chair professor, also the vice dean of the School of Microelectroincs. He has been served as the TPC Member of ISSCC and ASSCC, also elected as the Co-chair of the Circuit & System Division and the Fellow of the CIE. His research fields include information security chips, base-band processing technologies for wireless communication; mixed-signal IC designs and ultra-low power IC Methodology. Prof. Zeng has over 200 academic papers in publication and applied for 120 patents.

Fumin Zhang

Hong Kong University of Science and Technology

Title: Learning and Predicting Human Intentions and Actions for Autonomy

Abstract:  One of the most challenging problems for robot to effectively interact with human is the lack of predictive models for human Intentions and actions. The research community is making tremendous effort in developing learning algorithms that may generate predictive models, which leads to feedback control methods for robots that are able to adapt to individual differences and respect human comfort. This talk will focus on how to recognize and predict human intentions and actions when a robot may have an opportunity to trigger reactions from a human subject repeatedly. We will discuss models for human pointing motion, human feature detection, and a class of expert-based learning algorithms that have been verified by human robot interaction in experiments. We develop the Georgia Tech Miniature Autonomous Blimp (GT-MAB) as flying vehicles for indoor experiments that support safe interaction between human and flying robots.

Bio:  Dr.‬ Fumin ZHANG‬ is Chair Professor and Director of the Cheng Kar-Shun Robotics Institute at the Hong Kong University of Science and Technology. He is also Dean’s Professor adjunct in the School of Electrical and Computer Engineering at the Georgia Institute of Technology. He received a PhD degree in 2004 from the University of Maryland (College Park) in Electrical Engineering and held a postdoctoral position in Princeton University from 2004 to 2007. His research interests include mobile sensor networks, maritime robotics, control systems, and theoretical foundations for cyber-physical systems. He received the NSF CAREER Award in September 2009 and the ONR Young Investigator Program Award in April 2010. He is currently serving as the co-chair for the IEEE RAS Technical Committee on Marine Robotics, associate editors for IEEE Transactions on Automatic Control, and IEEE Transactions on Control of Networked Systems, IEEE Journal of Oceanic Engineering, and International Journal of Robotics Research.‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬

Hanwang Zhang

School of Computer Science and Engineering, Nanyang Technological University

Title: Recent Progress on Causality in Computer Vision

Abstract:  Causal intervention is a practical approach for achieving unbiased (debiasing) and out-of-domain (OOD) generalization in classification. In recent years of visual causality research, the construction of causal graphs has often relied heavily on the researcher's own understanding of the problem, resulting in subjectivity and controversy. This report will explore: 1) how to objectively construct causal graphs, 2) the tangible benefits that such objective causal graphs can bring, and 3) based on some comparisons, summarize the current shortcomings of visual causal interventions.

Bio:  Hanwang Zhang is an Associate Professor at Nanyang Technological University's School of Computer Science and Engineering. His research interests include Computer Vision, Natural Language Processing, Causal Inference, and their combinations. Due to his contribution in applied causality, he has received numerous awards including the Singapore President Award Young Scientist 2021, IEEE AI’s-10-To-Watch 2020, Alibaba Innovative Research Award 2019, Nanyang Assistant Professorship 2018, and several best paper awards.

Yue Zhang

Westlake University

Title: Research on Reasoning and Generalization of Large Models

Abstract:  In this talk, I will discuss linguistic reasoning, and the capabilities of formal logic reasoning for large langauge models (LLMs). I will discuss the difficulty of learning formal reasoning from empirical risk minimization, and discuss a perspective to this problem from causal learning theory. I will discuss causal features and confounders, and show how learning confounders can lead to low out-of-distribution generalization performance. Then I will discuss two general methods to address the issue, including a data-centric method and a model-centric method. For the former, I will introduce several methods for counterfactual data augmentation. For the latter, I will introduce methods to integrate causal features for machine translation.

Bio:  Yue Zhang is a tenured Professor at Westlake University. His research interests include NLP and its underlying machine learning algorithms. His major contributions to the field include psycholinguistically motivated machine learning algorithm, learning-guided beam search for structured prediction, pioneering neural NLP models including graph LSTM, and OOD generalization for NLP. He authored the Cambridge University Press book ``Natural Language Processing -- a Machine Learning Perspective''. He is the PC co-chair for CCL 2020 and EMNLP 2022, and action editor for Transactios for ACL. He also served as associate editor for IEEE/ACM Transactions of Audio Speech and Language Processing (TASLP), ACM Transactions on Asian and Low-Resource Languages (TALLIP), IEEE Transactions on Big Data (TBD) and Computer, Speech and Language (CSL). He won the best paper awards of IALP 2017 and COLING 2018, best paper honorable mention of SemEval 2020, and best paper nomination for ACL 2018 and ACL 2023.

Bolei Zhou

Computer Science Department, University of California, Los Angeles

Title: Human-in-the-loop Learning for Embodied AI

Abstract:  Embodied AI, which perceives and acts in the physical surroundings, has been deployed in many real-world applications from autonomous driving to household robots. However, it remains challenging to ensure AI safety and alignment with human intents and expectation. In this talk, I will show that human-in-the-loop learning can not only facilitate efficient training of the embodied AI for safe driving but also bring interpretable human-AI shared control for locomotion. Then I will talk about visuomotor policy pre-training conditioned on human actions from hours of uncurated YouTube videos. Finally, I will briefly introduce our on-going effort for building an open-source driving simulator called MetaDriverse for generalizable embodied AI, by incorporating a massive number of real-world scenarios and learning to generate novel ones.

Bio:  Bolei Zhou is an Assistant Professor in the Computer Science Department at the University of California, Los Angeles (UCLA). He earned his Ph.D. from MIT in 2018. His research interest lies at the intersection of computer vision and machine autonomy, focusing on enabling interpretable human-AI interaction. He has developed many widely used neural network interpretation methods such as CAM and Network Dissection, as well as computer vision benchmarks Places and ADE20K. He has been area chair for CVPR, ECCV, ICCV, and AAAI. He received MIT Tech Review's Innovators under 35 in Asia-Pacific Award.

Jun Zhou

School of Information and Communication Engineering, University of Electronic Science and Technology of China

Title: Energy-Efficient Domain-Specific AI Processor Design for AIoT Applications

Abstract:  This talk introduces two energy-efficient domain-specific AI processors: one for visual object detection & tracking applications and the other for wearable health monitoring applications.The two processors exploit diverse domain-specific features through algorithm-hardware co-optimization to achieve high energy efficiency while maintaining high accuracy and certain flexibility for domain-specific algorithms. They are suitable for AI processing on energy and resource constrained AIoT devices.

Bio:  Jun Zhou is currently a Professor at UESTC. His major research interest is algorithm and processor co-design for energy-efficient intelligent sensing. Before joining UESTC, he has worked as Research Scientist and Principle Investigator at IMEC Netherlands and Institute of Microelectronics A*STAR Singapore from 2008 to 2017. He has published more than 100 papers in prestigious conferences and journals, including ISSCC, JSSC, DAC, CICC, TBioCAS, and TCAS-I. He has served or is serving as A-SSCC TPC Sub-Committee Chair of Digital Circuits & Systems, Guest Editor of JSSC, Associate Editor of TBioCAS and Associate Editor of TVLSI. He has also served as a TPC/OC Member for a number of IEEE conferences, including DAC, SOCC, ICCD, and ISCAS.

S Kevin Zhou

University of Science and Technology of China

Title: Generative AI for Medical Imaging

Abstract:  "Medical images are widely used in clinical decision making and artificial intelligence (AI) technologies are commonly utilized in medical imaging and image analysis. In this talk, we address the aspect of medical imaging and present an overview of how AI facilitates the generation of medical image through recovery and synthesis. Medical image recovery attempts to recover the original image under adverse imaging conditions, such as metal artifacts, slow acquisition time, etc. Medical image synthesis attempts to synthesize, from an acquired image under current conditions, a novel image under different conditions. We will cover three neural approaches: (i) Dual domain network (DuDoNet) for metal artifact reduction in CT via joint learning in both sinogram and image domains and MR image reconstruction from undersampled k-space data via joint and recurrent learning in both frequency and image domains; (ii) Causal image synthesis (CIS) for counterfactually synthesizing MR brain images in 3D via the nontrivial leverage of a causal graph and 3D StyleGAN; and (iii) Unified multimodal image synthesis (UMIS) for imputing missing MR images of multiple modalities from any combination of available ones with a single unified model. Our recovery and synthesis approaches leverage deep neural networks as cores, integrate specific domain knowledge, and achieve high quality images."

Bio:  Prof. S. Kevin Zhou obtained his PhD degree from University of Maryland, College Park and currently is a Distinguished Professor and Founding Executive Dean of School of Biomedical Engineering, University of Science and Technology of China (USTC). He was a Principal Expert of AI and a Senior R&D Director at Siemens Healthineers Research. Dr. Zhou has published 260+ book chapters and peer-reviewed journal and conference papers, registered 150+ granted patents, and written and edited 6 research monographs. The two recent books he led the edition are entitled "Deep Learning for Medical Image Analysis, SK Zhou, H Greenspan, DG Shen (Eds.)" and "Handbook of Medical Image Computing and Computer Assisted Intervention, SK Zhou, D Rueckert, G Fichtinger (Eds.)". He has won multiple awards including R&D 100 Award (Oscar of Invention), Siemens Inventor of the Year, UMD ECE Distinguished Alumni Award, BMEF Editor of the Year, Finalist Paper for MICCAI Young Scientist Award (twice). He has been a program co-chair for MICCAI 2020 conference, an associate editor for IEEE Trans. Medical Imaging (TMI), IEEE Trans. Pattern Analysis and Machine Intelligence (TPAMI), and Medical Image Analysis, and an area chair for AAAI, CVPR, ICCV, MICCAI, and NeurIPS. He has been elected as a treasurer and board member of the MICCAI Society, an advisory board member of MONAI (Medical Open Network for AI), and a fellow of AIMBE, IAMBE, IEEE, MICCAI, and NAI.

Feng Zhu

College of Pharmaceutical Sciences, Zhejiang University

Title: Computer-assisted drug target discovery research

Abstract:  This presentation will report the latest research progress of the group in the direction of computer-aided drug target discovery.

Bio:  Mr. Zhu Feng is a Distinguished Professor of Zhejiang University, Editor-in-Chief of Elsevier Publishing Group's Comp Biol Med (IF=7.7), and Associate Editor of American Chemical Society's J Chem Inf Model (IF=5.6) in East Asia. His research work has been selected as one of the "China's 100 Most Influential International Academic Papers" by CITIC, Ministry of Science and Technology, and one of the "Top 10 Progresses in Bioinformatics in China in 2020". In the past five years, he has published more than 70 papers in Nature Protoc, Nucleic Acids Res, Adv Sci, Anal Chem, Engineering, Acta Pharmacol Sin, Mol Cell Proteomics, J Mol Biol, etc.,

Jun Zhu

Computer Science Department, Tsinghua University

Title: Large-scale Diffusion Models for Multimodal Generation

Abstract:  Diffusion probabilitistic models have shown great success on generating visual data (e.g., images). In this talk, I will present some recent progress on large-scale diffusion models. The first part will present the UniDiffuser for fitting any cross-modal generation task via a single transformer network. Then, I will present progress on efficient text-to-3D generation and text-to-video generation based on a pre-trained 2D diffusion model.s.

Bio:  Dr. Jun Zhu is a Bosch AI Professor at the Department of Computer Science and Technology in Tsinghua University and an IEEE Fellow. He was an Adjunct Faculty at the Machine Learning Department in Carnegie Mellon University (CMU) from 2015 to 2018. Dr. Zhu received his B.E and Ph.D. in Computer Science from Tsinghua in 2005 and 2009, respectively. Before joining Tsinghua in 2011, he did post-doctoral research in CMU. His research interest lies in machine learning and applications in text and image analysis. Dr. Zhu has published over 100 papers in the prestigious conferences and journals. He is an associate editor-in-chief for IEEE Trans. on PAMI. He served as senior area chairs for ICML, NeurIPS, and ICLR. He was a local co-chair of ICML 2014. He is a recipient of several awards, including ICLR Outstanding Paper Award, IEEE CoG Best Paper Award, XPlorer Prize, IEEE Intelligent Systems "AI's 10 to Watch" Award, MIT TR35 China, CCF Young Scientist Award, and CCF first-class Natural Science Award. His team has won several first-place awards in international competitions, including all the three tasks in NeurIPS 2017 adversarial attack and defense for deep learning and the intelligent decision task in ViZDoom 2018.

Chengqing Zong

Institute of Automation, Chinese Academy of Sciences

Title: Performance Analysis and Application Prospects of Large Language Model

Abstract:  The large language model (LLM) represented by ChatGPT has attracted much attention in the past half-year. Its good performance in various fields and tasks has shown people the hope of achieving artificial general intelligence. However, what are the differences in performance compared to the models specically trained on specific tasks? Is ChatGPT a good foundation model for all taks? And what are the future development prospects of the LLM? This report will present the results of specific comparative analysis and introduce some related work carried out by our research group. Through the introduction and discussion of these specific tasks, I hope to predict and prospect the future development of natural language processing technology.

Bio:  Chengqing Zong received his Ph.D. degree from the Institute of Computing Technology, Chinese Academy of Sciences (CAS), in March 1998. From May 1998 to April 2000, he worked at the National Laboratory of Pattern Recognition (NLPR), Institute of Automation, CAS, as a post-doctor research fellow, and he has been working with NLPR since finishing his post-doctoral program. In 1999 and 2001, Dr. Zong twice worked in the Advanced Telecommunications Research Institute International (ATR) of Japan, as a guest researcher. In 2004, he visited CLIPS-IMAG, France, as visiting scholar. Now he is a professor, IEEE Fellow, ACL Fellow, AAIA Fellow, CAAI Fellow and CCF Fellow. His research interests include machine translation, text data mining and cognitive language computing. He has authored/co-authored more than 200 papers. His book Statistical Natural Language Processing is very popular in NLP community of China, and the book Text Data Mining has received widespread attention. He served the top-tier international conference ACL 2015 as PC Co-Chair and ACL 2021 as General Conference Chair, and also the Asian Federation on Natural Language Processing (AFNLP) as president from 2019 to 2021. Now he is Vice-President-elect of Association for Computational Linguistics.