报告题目：Two Accelerated Non-backtracking PageRank Algorithms for Large-scale Networks
报告时间：2025年4月25日 15:00-16:00
报告地点：6教南楼 528
摘要：Non-backtracking PageRank is a variation of Google’s PageRank, which is based on non-backtracking random walk. However, if the number of dangling nodes of a graph is large, the non-backtracking PageRank algorithm may suffer from huge memory requirements and heavy computational costs. In this talk, the authors first consider how to compute the non-backtracking PageRank vector efficiently by using the Jacobi iteration, and then propose two strategies to speed up the computation of non-backtracking PageRank, in which we add some edges to a graph in a randomized and a fixed way, respectively. The advantages of the proposed algorithms are two-fold. First, the sizes of the matrix computation problems are much smaller than that of the original one. Second, there is no kronecker product in the involved non-backtracking edge matrices, and the structures of the non-backtracking PageRank problems are greatly simplified.
报告人简介：吴钢，中国矿业大学数学学院教授、博士生导师；江苏省“333工程”中青年科学技术带头人，江苏省“青蓝工程”中青年学术带头人，江苏省计算数学学会副理事长。主要研究方向：数值代数、数据科学中的大规模矩阵计算；数据挖掘与统计计算；大数据相关快速算法与理论；机器学习中的快速算法与优化；人工智能的数学理论与快速算法；先后主持国家自然科学基金项目、江苏省省自然科学基金项目多项，已在国际知名杂志，如：SIAM Journal on Numerical Analysis, SIAM Journal on Matrix Analysis and Applications, SIAM Journal on Scientific Computing, IMA Journal of Numerical Analysis, Pattern Recognition, Machine Learning等期刊发表学术论文多篇。
