• EdgeSketch is a compact graph sketch enabling efficient analysis of massive graph streams. • Provides unbiased estimators for key graph properties with controllable variance. • Supports direct implementation of graph algorithms on the sketch, eliminating need for full graph. • Constructed in a single pass over edge stream, fully streaming, no offline preprocessing. • Evaluated on community detection (Louvain) and graph reconstruction via node similarity. • Outperforms lossless representations and prior sketches in memory usage and runtime while maintaining accuracy.
Article Summaries:
- EdgeSketch is a new compact graph representation designed for streaming analysis of large‑scale graph data. The method constructs a sketch in a single pass over an edge stream, producing unbiased estimators for key graph properties with controllable variance. Unlike traditional lossless storage, EdgeSketch enables direct execution of graph algorithms on the sketch, eliminating the need to reconstruct the full graph. The authors evaluate the approach on community detection (Louvain method) and graph reconstruction via node similarity, reporting significant memory savings and runtime gains compared to both lossless representations and prior sketching techniques while preserving accuracy.
Sources: