2018/01/01 by Fei Jiang, Lifang He, Jiang, Fei +9
Computer Science · Neuroscience · Physics and Astronomy · #Advanced Graph Neural Networks #Complex Network Analysis Techniques #Data Structures and Algorithms (cs.DS) #FOS: Computer and information sciences #Functional Brain Connectivity Studies #Social and Information Networks (cs.SI)
paper · pdf · doi:10.48550/arxiv.1801.05855
openalex publication_date 2018/01/01 · openalex created_date 2019/01/11 · openalex updated_date 2026/07/28
Graph embedding has been proven to be efficient and effective in facilitating\ngraph analysis. In this paper, we present a novel spectral framework called\nNOn-Backtracking Embedding (NOBE), which offers a new perspective that\norganizes graph data at a deep level by tracking the flow traversing on the\nedges with backtracking prohibited. Further, by analyzing the non-backtracking\nprocess, a technique called graph approximation is devised, which provides a\nchannel to transform the spectral decomposition on an edge-to-edge matrix to\nthat on a node-to-node matrix. Theoretical guarantees are provided by bounding\nthe difference between the corresponding eigenvalues of the original graph and\nits graph approximation. Extensive experiments conducted on various real-world\nnetworks demonstrate the efficacy of our methods on both macroscopic and\nmicroscopic levels, including clustering and structural hole spanner detection.\n