Fast and Attributed Change Detection on Dynamic Graphs with Density of States

How can we detect traffic disturbances from international flight transportation logs, or changes to collaboration dynamics in academic networks? These problems be formulated as detecting anomalous change points a dynamic graph. Current solutions do not scale well large real world graphs, lack robustness amount of node additions / deletions and overlook attributes. To address these limitations, propose novel spectral method: Scalable Change Point Detection (SCPD). SCPD generates an embedding for each graph snapshot by efficiently approximating the distribution Laplacian spectrum at step. also capture shifts attributes tracking correlations between eigenvectors. Through extensive experiments using synthetic data, show that (a) achieves state-of-the-art performance, (b) is significantly faster than methods easily process millions edges few CPU minutes, (c) effectively tackle quantity attributes, (d) discovers interesting events graphs. Code publicly available https://github.com/shenyangHuang/SCPD.git .