講座主題：Graph and Graph Neural Networks: an Algebraic Topology Perspective
專(zhuān)家名稱(chēng)：Jian Yu (喻堅(jiān))
工作單位：奧克蘭理工大學(xué)
講座時(shí)間：2026年01月05日 09:30-10:30
講座地點(diǎn)：科技館4306
主辦單位：煙臺(tái)大學(xué)計(jì)算機(jī)與控制工程學(xué)院
內(nèi)容摘要：
In this talk, we first examine the graph as a discrete structure from an algebraic topology perspective and put graphs into the context of simplicial complexes and chain complexes. We then connect the matrices used to encode graphs, including incidence matrix, adjacency matrix, and graph Laplacian matrix, to concepts of gradient, divergence, and the boundary and co-boundary operators in chain complexes. Based on that, we examine the function of the popular GCN (Graph Convolutional neural network) layer and relate it to the heat diffusion differential equation. finally, we introduce our recent works on graph motifs based bipartite graph link prediction and its application in recommender systems.
主講人介紹：
Prof. Dr. Jian Yu is a full professor in the Department of Computer and Information Sciences, Auckland University of Technology. He is currently the director of the Ubiquitous and Intelligent Web Computing Research Lab (UbiWeb). He holds a PhD degree in Computer Software and Theory from Peking University. His current research interests include deep learning for recommender systems, graph neural networks, complex networks, Web and ubiquitous computing, and service-oriented computing. Prof Yu is the recipient of the 2025 New Zealand-China Scientists Exchange Program. He is Associate Editor for some top journals including IEEE Transactions on Services Computing (CORE A*-Top 6%), and has organized over 10 special issues and served as PC Member for over 100 international conferences. He has over 160 publications.
