PTCL trains a dynamic graph classifier with only final-timestamp labels by using a variational EM loop and a temporal curriculum that weights pseudo-labels by closeness to the final label.
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
-
PTCL: Pseudo-Label Temporal Curriculum Learning for Label-Limited Dynamic Graph
PTCL trains a dynamic graph classifier with only final-timestamp labels by using a variational EM loop and a temporal curriculum that weights pseudo-labels by closeness to the final label.