Baojian Zhou
Young Associate Researcher, School of Big Data, Fudan University
Baojian Zhou is a Young Associate Researcher at School of Big Data, Fudan University.He received his Ph.D. degree from the State University of New York at Albany in 2020, followed by a postdoctoral fellowship at Stony Brook University (2020-2021). He has long been engaged in theoretical research work on large-scale graph machine learning and data mining, especially on dynamic graph representation learning and acceleration algorithms on graphs. In recent years, he has achieved a number of research results in the research directions of anomaly subgraph detection, graph structure constraint optimization, graph machine learning, and anomaly detection for dynamic graph representation learning on large-scale graph data mining. The research results have been published in more than 20 papers in international data mining and machine learning conferences, such as ICML, NeurIPS, KDD, IJCAI, AAAI, ICDM, CIKM, and TKDE.
Topic
Efficient local computation and optimization on large-scale graphs
Graph Diffusion Equations (GDEs) are fundamental tools for modeling graph data and have been successfully applied to many graph learning tasks, including local clustering, semi-supervised learning, node embedding, graph neural network (GNN) training, and many other applications. In this presentation, I will present recent advances in local iterative methods for solving graph diffusion problems. I will introduce a new framework, the Locally Evolving Set Process (LESP), which effectively localizes standard iterative solvers. The method significantly improves the speed of diffusion vector computation with sub-linear runtime complexity, reflecting the performance of the algorithm in real-world applications. The framework exploits the localization property of diffusion vectors to provide significant computational savings and is particularly suitable for large-scale dynamic graphs on GPUs. A number of open problems and future research directions will also be discussed in the presentation.
