VALSE Webinar 20240110-02期 总第336期 图数据小样本学习 Few shot learning on grap ...

报告嘉宾：姚权铭 (清华大学)

报告题目：Few-shot learning from biomedical networks

报告嘉宾：方元 (新加坡管理大学)

报告题目：Would prompt work for graph learning? An exploration of few-shot learning on graphs

报告嘉宾：姚权铭 (清华大学)

报告时间：2024年1月10日 (星期三)晚上20:00 (北京时间)

报告题目：Few-shot learning from biomedical networks

报告人简介：

姚权铭目前是清华大学电子工程系助理教授，国家高层次青年人才计划入选者。于香港科技大学计算机系取得博士学位，后于第四范式担任首席研究员，创建和领导机器学习研究团队。主要研究方向为机器学习，特别是结构化数据元学习方法。发表顶级论文80余篇，包括Nature Comp. Sci./ Nat. Com./ JMLR/ IEEE TPAMI/ ICML/ NeurIPS/ ICLR等，总被引8000余次。其中抗噪标签算法“Co-teaching”是鲁棒学习领域的里程碑；小样本学习综述是CSUR近五年来最高被引论文；自动化图学习方法 (TPAMI 2023等)蝉联Open Graph Benchmark榜单第一名；基于医药网络解决新药物互反应的相关工作，刊载于Nature子刊。担任ICML、NeurIPS和ICLR会议领域主席，期刊Neural Network和Machine Learning编委。荣获国内外诸多知名奖项，包括国际神经网络学会早期成就奖、香港科学会优秀青年科学家、Google全球博士奖等，同时入选全球Top 50华人AI青年学者榜、福布斯30Under30精英榜与全球Top 2%科学家。

个人主页：

http://cvlab.cse.msu.edu

报告摘要：

Accurately predicting drug-drug interactions (DDI)for emerging drugs, which offer possibilities for treating and alleviating diseases, with computational methods can improve patient care and contribute to efficient drug development. However, existing methods require large amounts of known DDI information, which is scarce for emerging drugs. In this talk, we present EmerGNN, a graph neural network (GNN)that can effectively predict interactions for emerging drugs by leveraging the rich information in biomedical networks. EmerGNN learns pairwise representations of drugs by extracting the paths between drug pairs, propagating information from one drug to the other, and incorporating the relevant biomedical concepts on the paths. Overall, EmerGNN has higher accuracy than existing approaches in predicting interactions for emerging drugs and can identify the most relevant information on the biomedical network. Related works have recently been accepted to Nature Computational Science/ Nature Communication/ IEEE TPAMI.

参考文献：

[1] Emerging Drug Interaction Prediction Enabled by Flow-based Graph Neural Network with Biomedical Network. Nature Computational Science. 2023

[2] Accurate and Interpretable Drug-drug Interaction Prediction Enabled by Knowledge Subgraph Learning. Nature Communications (Medicine). 2023

[3] Bilinear Scoring Function Search for Knowledge Graph Learning. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). 2023

报告嘉宾：方元 (新加坡管理大学)

报告时间：2024年1月10日 (星期三)晚上20:30 (北京时间)

报告题目：Would prompt work for graph learning? An exploration of few-shot learning on graphs

报告人简介：

Dr. Yuan Fang is a tenure-track Assistant Professor at the School of Computing and Information Systems at Singapore Management University (SMU). He was previously a data scientist at DBS Bank and a research scientist at A*STAR. His research interests revolve around graph-based learning and its applications in recommender systems, social network analysis, and bioinformatics. Dr. Fang has published over 60 papers in leading conferences and journals, and his work has been featured in the "Best Papers of VLDB13" special issue of the VLDB Journal. He serves as an associate editor of Frontiers of Computer Science, and as an organization committee member as well as PC/ SPC member/ AC for various international conferences.

个人主页：

https://www.yfang.site/

报告摘要：

Graph structures are prevalent across a variety of fields, including social networks, e-commerce, transportation, and biological systems. Within these graphs, numerous analytical and mining tasks can be identified, often aligning with link prediction, node classification, or graph classification. Graph Neural Networks (GNNs) have achieved significant success in these applications, primarily due to their capability to learn powerful graph representations. Nevertheless, the efficacy of GNNs often depends on the availability of labeled data, without which their performance may be compromised. This talk seeks to delve into learning paradigms that diverge from traditional supervised learning, in the context of few-shot learning on graphs. In particular, drawing inspiration from recent progress in language modeling, we pose an intriguing question: Can prompt-based learning be effectively adapted for graph data? Our talk will begin with a comprehensive overview of few-shot learning methodologies on graphs, followed by highlighting some of our representative works in this area.

参考文献：

[1] Xingtong Yu, Yuan Fang, Zemin Liu and Xinming Zhang. HGPrompt: Bridging Homogeneous and Heterogeneous Graphs for Few-shot Prompt Learning. Accepted by AAAI2024.

[2] Zhihao Wen and Yuan Fang. Augmenting Low-Resource Text Classification with Graph-Grounded Pre-training and Prompting. In SIGIR 2023, pp. 506--516.

[3] Zemin Liu, Xingtong Yu, Yuan Fang and Xinming Zhang. GraphPrompt: Unifying Pre-Training and Downstream Tasks for Graph Neural Networks. In TheWebConf 2023, pp. 417--428.

[4] Jiawei Liu, Cheng Yang, Zhiyuan Lu, Junze Chen, Yibo Li, Mengmei Zhang, Ting Bai, Yuan Fang, Lichao Sun, Philip S. Yu, Chuan Shi. Towards Graph Foundation Models: A Survey and Beyond. In arXiv:2310.11829.

主持人：王鑫 (清华大学)

主持人简介：

王鑫，男，博士，清华大学计算机系副研究员，国家优秀青年科学基金获得者，浙江大学计算机科学与技术学士、博士，加拿大西蒙弗雷泽大学计算科学博士。中国计算机学会多媒体专业技术委员会副秘书长，清华大学博士后校友会秘书处副秘书长。主要研究方向为多媒体智能计算，大数据分析，机器学习，在TPAMI, TKDE, TOIS, TMM, ICML, NeurIPS, KDD, WWW, ACM Multimedia, SIGIR等相关领域顶级期刊及会议上发表论文150余篇。作为项目/ 课题负责人承担国家重点研发专项、国家自然科学基金等项目，获ACM中国新星奖、IEEE TCMC Rising Star Award、达摩院青橙奖、教育部自然科学一等奖。

个人主页：

https://mn.cs.tsinghua.edu.cn/xinwang/

特别鸣谢本次Webinar主要组织者：

主办AC：王鑫 (清华大学)

协办AC：郑乾 (浙江大学)