Landmark-Based Methods Outperform Random Walks for Large Graph Distance Estimation
August 19, 2026
Landmark-based approaches using HyperLogLog probabilistic data structures provide superior memory efficiency and accuracy for estimating average distances in large-scale static graphs. The Eppstein-Wang algorithm offers higher accuracy with lower computation time than the Size Estimation Framework (SEF) or Random Walk sampling.
HOW THIS AFFECTS YOU
●
builderUse the EW algorithm or SEF with HyperLogLog to estimate graph metrics when main memory is constrained.
●
researcherThe findings highlight the scalability limits of Random Walk sampling for small-scale graph node requirements.