Scalable Graph Condensation with Evolving Capabilities

Shengbo Gong, Mohammad Hashemi, Juntong Ni, Carl Yang, Wei Jin · 2025

Read Scalable Graph Condensation with Evolving Capabilities by Shengbo Gong, Mohammad Hashemi, Juntong Ni, Carl Yang, Wei Jin (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-17614.txt.

Description

Shengbo Gong and co-authors shrink big network data in a 2025 study so learning stays fast as the network keeps growing. Graph condensation boils a large graph down to a small one that trains models nearly as well. Old methods assume a fixed data set and must redo the costly boil when new data lands.

Their method clusters nodes by class on summed features and reuses past cluster centers as the graph expands, so updates stay cheap. Tests show quality on par with top rivals at a tiny fraction of the time, near a thousand-fold faster on large data. Evolving graphs get a condenser that evolves with them.

Catalog record

CREATORShengbo Gong, Mohammad Hashemi, Juntong Ni, Carl Yang, Wei Jin
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE88 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-17614

Browse the full library · About the Tower of Babel