Knowledge distillation via adaptive meta-learning for graph neural network

Tiesunlong Shen, Jin Wang, Xuejie Zhang

Abstract: Highlights•This study proposed an adaptive meta-learning method for knowledge distillation in graph neural networks.•Teacher model updates parameters based on student's feedback, improving knowledge transfer.•Local structure preservation introduced to avoid over-smoothing in graph neural networks.•Experimental results show superior performance on multiple datasets compared to baselines.