Highlight Paper Speakers
Ordered List by Surname Alphabetical Order:
l Ming Li
l Yuhua Xu
|
|
Linwei Chen Beijing Institute of Technology, China PhD Candidate |
|
|
Longyang Huang Shanghai Jiao Tong University, China PhD Candidate |
|
|
Xiaojun Jia Nanyang Technological University, Singapore PhD |
|
|
Ming Li Zhejiang Normal University, China PhD & Professor |
|
|
Guangzhou Laboratory, China PhD & Professor |
|
|
Xiaorui Su Harvard Medical School, Boston, USA PhD |
|
|
Handing Wang Xidian University, Xi'an, China PhD & Professor |
|
|
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, China PhD & Professor |
|
|
China University of Mining and Technology, Xuzhou, China PhD & Professor |
|
|
Yuhua Xu Donghua University, Shanghai, China PhD |
|
|
Wen Zhang Huazhong Agricultural University, Wuhan, China PhD & Professor |
Frequency-aware Feature Fusion for Dense Image Prediction
Linwei Chen, PhD Candidate
Beijing Institute of Technology, China
Bio-Sketch: Linwei Chen
is a PhD candidate from the School of Computer
Science, Beijing Institute of Technology. His primary research interests lie in
computer vision, with a focus on areas including image segmentation, object
detection, low-light image enhancement and recognition, and image generation.
To date, he has published over ten papers, many of which are as the first
author in top-tier international journals and conferences in computer vision,
such as TPAMI, IJCV, CVPR, ICLR, and ISPRS. In terms of his contributions to
the academic community, he serves as a reviewer for numerous journals and
conferences, including IJCV, TIP, CVPR, ICCV, NeurIPS, and AAAI. He was
recognized as an "Outstanding Reviewer" at the British Machine Vision
Conference (BMVC) for his professional expertise and contributions.
His outstanding achievements is
Proposing the integration of three modules—an adaptive low-pass filter, an offset generator, and an adaptive high-pass filter—from a frequency-domain perspective. This approach effectively addresses the challenges of intra-class inconsistency and boundary displacement in dense image prediction, achieving state-of-the-art performance on multi-tasks such as semantic segmentation and object detection (IEEE Transactions on Pattern Analysis and Machine Intelligence (2024): Frequency-aware Feature Fusion for Dense Image Prediction).
Efficient Offline Reinforcement Learning with Relaxed Conservatism
Longyang Huang, PhD Candidate
Shanghai Jiao Tong University, China
Bio-Sketch: Longyang Huang
is a PhD Candidate at Shanghai Jiao Tong University. He is dedicated to fundamental
theoretical research in control theory and artificial intelligence, with core
research directions including reinforcement learning and optimal control, as
well as learning-based control and decision theory. His papers have been
published in journals such as IEEE TPAMI and IEEE TNNLS, and he serves as a
reviewer for journals like IEEE TAC and IEEE TNNLS.
His outstanding achievements is
Proposing a simple and efficient offline RL framework with relaxed conservatism. This framework addresses the conservatism of learning policies by learning Q-functions close to real-world values under the learning policy (IEEE Transactions on Pattern Analysis and Machine Intelligence (2024): Efficient Offline Reinforcement Learning with Relaxed Conservatism).
He has won the National Scholarship for Doctoral Students and the Jiangsu Provincial Excellent Master's Thesis Award.
Semantic-Aligned Adversarial Evolution Triangle for High-Transferability Vision-Language Attack
Xiaojun Jia, Dr.
Nanyang Technological University, Singapore
(Online)
Bio-Sketch: Xiaojun Jia, a PhD from the Chinese
Academy of Sciences, is currently conducting postdoctoral research at Nanyang
Technological University. His doctoral research directions cover computer
vision, adversarial attacks, adversarial training, reinforcement learning, and
other fields. From February to September 2023, Jia Xiaojun studied as a remote
visiting student at the Torr Vision Group, University of Oxford. From March
2022 to February 2023, he served as a research intern in the security
department of Alibaba Group; from May 2020 to February 2022, he was a research
intern at Tencent AI Laboratory. Currently,
his research focuses on large model security-related issues, including
jailbreak attacks against large language models (LLMs), adversarial
transferability of vision-language models (VLMs), etc.
His outstanding achievements is
Proposing a semantic-aligned adversarial evolution triangle framework, which systematically addresses the semantic gap problem in cross-modal attacks and achieves breakthrough improvements in attack transferability across multiple vision-language tasks such as image captioning and visual question answering (IEEE Transactions on Pattern Analysis and Machine Intelligence (2025): Semantic-Aligned Adversarial Evolution Triangle for High-Transferability Vision-Language Attack).
Are graph convolutional networks with random weights feasible?
Ming Li, Dr. Prof.
Zhejiang Normal University, China
Ming Li:
Ming Li, Ph.D. Supervisor, is recognized as a High-Level Leading Talent in the
universities of Zhejiang Province and is selected for the Zhejiang Provincial Qiantang
River Talent Program. He currently serves as the Associate Director of the Zhejiang
Provincial Key Laboratory of Intelligent Education Technology and Applications.
His editorial roles include Associate Editor or Editorial Board Member for
several prestigious journals, notably the Chinese Academy of Sciences Tier 1
TOP journal Neural Networks. He is also a member of the IEEE Task Force on
Learning For Graphs. Previously, he served as the Lead Guest Associate Editor
for the special issue "Deep Neural Networks for Graphs: Theory, Models,
Algorithms and Applications" in the IEEE Transactions on Neural Networks
and Learning Systems. He has published over 100 papers in internationally
authoritative journals such as IEEE TPAMI, AI, and IEEE TKDE, as well as at
CCF-A ranked conferences including ICML, NeurIPS, IJCAI, AAAI, and ICDE. He is
listed among the World's Top 2% Most Influential Scientists. His research has
garnered over 4,000 citations on Google Scholar, with an h-index of 32.
His outstanding highlight achievements is:
Proposing a novel model termed Graph Convolutional Networks with Random Weights (GCN-RW) by revising the convolutional layer with random filters and simultaneously adjusting the learning objective with regularized least squares loss. (IEEE Transactions on Pattern Analysis and Machine Intelligence (2023): Are Graph Convolutional Networks With Random Weights Feasible?).
Alleviating Batch Effects in Cell Type Deconvolution with SCCAF-D
Zhichao Miao, Dr. Prof.
Guangzhou Laboratory, China
His outstanding highlight achievements include:
1) Development of the cell deconvolution algorithm SCCAF-D (Nature Communications (2024): Alleviating Batch Effects in Cell Type Deconvolution with SCCAF-D).
2) Deciphering the evolutionary mechanisms of cross-species visual cortex cells (Nature Communications (2022): Identification of Visual Cortex Cell Types and Species Differences Using Single-Cell RNA Sequencing).
3) Identification of neural precursor cell populations in the primate hippocampus (Nature Neuroscience (2022): Single-Cell Transcriptomics of Adult Macaque Hippocampus Reveals Neural Precursor Cell Populations).
4) Development of high-throughput full-length single-cell sequencing automation technology (Nature Protocols (2021): High-Throughput Full-Length Single-Cell RNA-Seq Automation).
5) Initiation of the single-cell unsupervised annotation algorithm SCCAF (Nature Methods (2020): Putative Cell Type Discovery from Single-Cell Gene Expression Data).
6) Establishment of the gold standard kBET for single-cell batch effect correction evaluation (Nature Methods (2019): A Test Metric for Assessing Single-Cell RNA-Seq Batch Correction).
He currently serves as the editor-in-chief/member of the editorial board of multiple international authoritative journals such as Nucleic Acids Research.
Xiaorui Su, Dr.
Harvard Medical School, Boston, USA
(Online)
Bio-Sketch: Xiaorui Su
earned her PhD from the Chinese Academy of Sciences. She is currently
conducting postdoctoral research at Harvard Medical School. She, specializing
in graph representation learning models to advance the understanding of
biomedical knowledge graphs. Her research particularly focuses on elucidating
molecular interactions, including drug-drug interactions and drug-target
interactions, and has made significant progress in facilitating drug
repurposing and cancer-related gene identification. Currently, her research
emphasis has shifted toward developing knowledge graph-based agent systems,
aiming to enhance the reasoning capabilities of large language models in
complex medical question-answering scenarios to provide deeper and more
accurate insights for clinical decision-making.
Her outstanding achievements is
Exploring interpretable identification of cancer genes across biological networks via transformer-powered graph representation learning, which achieved high-precision prediction of cancer genes by integrating multi-omics data with the topological structures of biological interaction networks (Nature Biomedical Engineering (2025): Interpretable Identification of Cancer Genes across Biological Networks via Transformer-Powered Graph Representation Learning).
Optimizing Latent Variables in Integrating Transfer and Query Based Attack Framework
Handing Wang, Dr. Prof.
Xidian University, Xi'an, China
Bio-Sketch:
Handing Wang received the PhD degrees from Xidian University,
Xi'an, China, in 2015. She worked as a researcher in the Computer Department of
the University of Surrey from 2015 to 2018, and was selected as a
provincial-level young talent, and was selected as a national-level young
talent in 2020. She is currently a professor with School of Artificial
Intelligence, Xidian University, Xi'an, China. Her research interests include
nature-inspired computation, multiobjective optimization, surrogate-assisted
evolutionary optimization, Trustworthy AI, and real-world problems.
Her outstanding achievements include:
1) Proposing an algorithm to optimize latent variables in the framework of integrated migration and query-based attack (IEEE Transactions on Pattern Analysis and Machine Intelligence (2024): Optimizing Latent Variables in Integrating Transfer and Query Based Attack Framework).
2) Developing a pre-training optimization model algorithm for zero-shot black box optimization (The Thirty-Eight Annual Conference on Neural Information Processing Systems (NeuRIPS 2024): Pretrained Optimization Model for Zero-Shot Black Box Optimization).
She is an Associate Editor of IEEE Transactions on Evolutionary Computing, Evolutionary Computing, Swarm and Evolutionary Computing, Complex & Intelligent Systems and Memetic Computing. She is a member of the program committee of the top international conferences in the field of evolutionary computing, such as Genetic and Evolutionary Computing Conference, IEEE Congress of Evolutionary Computing and many other international conferences. She has long been a reviewer of many top international journals in the field of computational intelligence.
Shanshan Wang, Dr. Prof.
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, China
Bio-Sketch: Shanshan
Wang, a joint PhD graduate from Shanghai Jiao Tong University and the
University of Sydney, currently is a professor at the Paul C Lauterbur Research
Center, Chinese Academy of Sciences, where she
targets to develop novel adaptive learning methods and applications for fast
medical imaging and intelligent medical analysis. She is affiliated with the
National Key Laboratory of Medical Imaging Science and Technology
Systems/Lauterbur Research Center for Biomedical Imaging. Her research
interests include fast medical imaging with AI, bioinformatics, and medical
imaging analysis. She also got selected as the World's Top 2% of Scientists by
Elsevier and Stanford University, USA from 2020-2024. She
has led six national-level research projects, including “AI 2030” initiative
and key projects, NSFC Joint Fund and Excellent Young Scholar Fund.
She has published more than 100 high-quality papers in journals
such as Nature Communications and IEEE Transactions on Medical Imaging, and has
been authorized over ten invention patents (including multiple industrial
applications). Her work has earned her numerous prestigious awards, including
the First Prize for Technological Invention from the Wu Wenjun AI Science and
Technology Award (1/6), the OCSMRM Distinguished Research Award (1/1), First
Prizes for Technological Invention and Scientific Progress in Guangdong
Province (2/10), and the Guangdong Youth Science and Technology Award (1/1).
Her outstanding achievements include:
1) Proposing a fine-grained alignment algorithm based on masked contrastive learning (Nature Communications (2024): Enhancing Representation in Radiography-Reports Foundation Model: A Granular Alignment Algorithm Using Masked Contrastive Learning).
2) Developing an annotation-efficient intelligent segmentation framework (Nature Communications (2021): Annotation-Efficient Deep Learning for Automatic Medical Image Segmentation).
Currently, she serves as an Associate Editor for several internationally authoritative journals, including IEEE Transactions on Medical Imaging, IEEE Transactions on Computational Imaging, Magnetic Resonance in Medicine, Pattern Recognition, and Biomedical Signal Processing and Control.
Local Superior Soups: A Catalyst for Model Merging in Cross-Silo Federated Learning
Zehua Wang, Dr. Prof.
China University of Mining and Technology, Xuzhou, China
Bio-Sketch: Zehua Wang,
PhD, is currently a Distinguished Professor of Jiangsu Province, Researcher at
China University of Mining and Technology, Honorary Visiting Professor of the
Department of Electrical and Computer Engineering at the University of British
Columbia (UBC), core member of Blockchain@UBC, and Chief Scientist of multiple
blockchain and artificial intelligence innovation projects. He has long been
dedicated to the research of blockchain, decentralized privacy computing,
federated learning, and artificial intelligence systems, presiding over several
projects funded by the Natural Sciences and Engineering Research Council of
Canada (NSERC), and publishing more than 80 papers in internationally
authoritative journals and conferences such as IEEE and ACM.
His outstanding achievements is
Proposing the "Local Superior Soups" method by integrating model merging in cross-silo federated learning, effectively promoting model fusion in cross-silo federated learning (The Thirty-Eighth Annual Conference on Neural Information Processing Systems (NeurIPS 2024): Local Superior Soups: A Catalyst for Model Merging in Cross-Silo Federated Learning).
He is committed to promoting the practical implementation of blockchain technology in energy management, financial compliance, and personalized AI training. As a bridge between governments, universities, and industries, he has been awarded by the Parliament of Canada for his outstanding contributions to promoting digital inclusion.
Graph Multi-Convolution and Attention Pooling for Graph Classification
Yuhua Xu, Dr.
Donghua University, Shanghai, China
Bio-Sketch: Yuhua Xu is a Lecturer and Master's
Supervisor at the College of Computer Science and Technology, Donghua
University. Her primary research focuses on financial fraud detection and graph
machine learning, with related work published in top-tier international
academic journals such as IEEE TPAMI and IEEE TNNLS. And she is a reviewer for
several journals, including IEEE TKDE, IEEE TII, IEEE TNNLS, and IEEE TIFS.
Her outstanding achievements is
Proposing a graph classification method based on graph multi-convolution and attention pooling, which enhances feature representation capabilities through multi-scale feature capture and key region focusing (IEEE Transactions on Pattern Analysis and Machine Intelligence (2024): Graph multi-convolution and attention pooling for graph classification).
Overcoming Collaboration Barriers in Quantitative Trait Loci Analysis
Wen Zhang, Dr. Prof.
Huazhong Agricultural University, Wuhan, China
Bio-Sketch: Wen Zhang is a Professor, PhD
supervisor, and Vice Dean at the College of Informatics, Huazhong Agricultural
University. He is a Distinguished Member of the China Computer Federation
(CCF), served as the Chair of CCF YOCSEF Wuhan (2022–2023), and is a member of
the CCF Technical Committees on Bioinformatics and Computer Applications, as
well as a member of the Bioinformatics and Artificial Life Technical Committee
of the Chinese Association for Artificial Intelligence. He serves on the
editorial boards of Briefings in Bioinformatics, IEEE Journal of Biomedical and
Health Informatics, among others, and has been listed in Stanford University’s
“Top 2% of Scientists Worldwide.” His research focuses on AI-driven drug
discovery, particularly applying deep learning and graph neural networks to
disease target identification, lead compound screening, drug property
prediction, and molecular design. He has developed a series of original
algorithms and led the compilation of the 2022 China AI White Paper chapter on
Artificial Intelligence and Drug Discovery. In recent years, he has led
multiple National Natural Science Foundation projects and has published over
100 papers as first or corresponding author in journals such as Cell Genomics,
Advanced Science, and Cell Reports Methods, and at top-tier AI conferences
including ACL, AAAI, and IJCAI. His publications include over ten papers with
an impact factor above 10 and more than 50 in CCF-A/B venues, with a total of
approximately 6,000 citations on Google Scholar. As the first contributor, he
received the 2024 Wu Wenjun Artificial Intelligence Science and Technology
Award (Second Prize in Natural Science) for his work titled Graph Learning
Methods for Drug Discovery: Research and Applications.
His outstanding achievements include:
1) Overcoming collaboration barriers in quantitative trait loci analysis (Cell Genomics (2025): Overcoming Collaboration Barriers in Quantitative Trait Loci Analysis).
2) Proposing DeepInterAware, a deep interaction interface-aware network, to improve antigen-antibody interaction prediction from sequence data (Advanced Science (2025): DeepInterAware: Deep Interaction Interface-Aware Network for Improving Antigen-Antibody Interaction Prediction from Sequence Data).