OGCIL replays old-class embeddings generated by a conditional VAE and rejects unknown nodes via prototype-centered hypersphere boundaries, improving open-set graph class-incremental learning on five benchmarks.
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
-
Towards Effective Open-set Graph Class-incremental Learning
OGCIL replays old-class embeddings generated by a conditional VAE and rejects unknown nodes via prototype-centered hypersphere boundaries, improving open-set graph class-incremental learning on five benchmarks.