IPAL improves non-exemplar continual graph learning by combining prototype contrastive learning with PageRank-weighted prototypes, instance-prototype affinity distillation, and decision-boundary hard-example mining, outperforming prior methods on four node classification benchmarks.
Exemplar-free continual representation learning via learnable drift compensation
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
-
Instance-Prototype Affinity Learning for Non-Exemplar Continual Graph Learning
IPAL improves non-exemplar continual graph learning by combining prototype contrastive learning with PageRank-weighted prototypes, instance-prototype affinity distillation, and decision-boundary hard-example mining, outperforming prior methods on four node classification benchmarks.