This study evaluates whether spectral filtering can accelerate continuous subgraph matching (CSM) on dynamic graphs, finding that while lazy maintenance is ineffective, selective exact maintenance offers significant performance gains.
- Lazy maintenance of spectral bounds loses pruning power within four updates, making it infeasible where spectral pruning is most valuable.
- Exact maintenance is affordable when selective because pruning utility and recomputation cost are anti-correlated; hubs never prune, allowing microsecond-level updates on small neighborhoods.
- Integrated tests remove up to 51% of candidates or skip 47% of update enumerations across two engines, four real graphs, and 77 queries.
- The approach accelerates construction and list scans but does not affect adjacency-guided exploration intermediates, except in specific radius-stratified workloads where it was 748x faster.
The authors distill an intermediate-invariance methodology for evaluating CSM filters and release a reusable dynamic local-spectra index to support this efficient maintenance strategy.