Distilling Holistic Knowledge with Graph Neural Networks

TL;DR

This paper introduces a novel knowledge distillation method that leverages graph neural networks to integrate and distill holistic knowledge, combining individual and relational information for improved student network learning.

Contribution

It proposes a unified graph-based approach to distill holistic knowledge, capturing correlations between individual and relational knowledge in a single framework.

Findings

Effective on benchmark datasets

Outperforms existing KD methods

Validated through extensive ablation studies

Abstract

Knowledge Distillation (KD) aims at transferring knowledge from a larger well-optimized teacher network to a smaller learnable student network.Existing KD methods have mainly considered two types of knowledge, namely the individual knowledge and the relational knowledge. However, these two types of knowledge are usually modeled independently while the inherent correlations between them are largely ignored. It is critical for sufficient student network learning to integrate both individual knowledge and relational knowledge while reserving their inherent correlation. In this paper, we propose to distill the novel holistic knowledge based on an attributed graph constructed among instances. The holistic knowledge is represented as a unified graph-based embedding by aggregating individual knowledge from relational neighborhood samples with graph neural networks, the student network is learned by distilling the holistic knowledge in a contrastive manner. Extensive experiments and ablation studies are conducted on benchmark datasets, the results demonstrate the effectiveness of the proposed method. The code has been published in https://github.com/wyc-ruiker/HKD