A pseudo-labeling approach based on knowledge distillation for graph few-shot learning

Open Webpage

Zongqian Wu, Peng Zhou, Guoqiu Wen, Xiaofeng Zhu

Published: 01 Jan 2025, Last Modified: 05 Nov 2025Inf. Process. Manag. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Embedding transfer preserves structure from base classes to aid pseudo-labeling.•Pseudo-label improvement reduces noise and boosts performance on novel classes.•First to apply knowledge distillation to FSNC on graphs; validated on six datasets.