DyGC, the first framework for dynamic graph condensation, synthesizes a small temporal graph that preserves enough spatiotemporal structure to train dynamic GNNs with up to 1846 times speedup and around 96 percent fidelity.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Dynamic Graph Condensation
DyGC, the first framework for dynamic graph condensation, synthesizes a small temporal graph that preserves enough spatiotemporal structure to train dynamic GNNs with up to 1846 times speedup and around 96 percent fidelity.