Deep graph clustering via aligning representation learning

Zhikui Chen, Lifang Li, Xu Zhang, Han Wang

Abstract: Highlights•A novel contrastive deep graph clustering method termed Aligning Representation Learning Network (ARLN) to avoid representation collapse problem.•ARLN selects positive/negative pairs in the original view without data augmentation, thus avoiding more computational costs.•Instead of merely utilizing contrastive learning in representation learning, our proposed contrastive strategy benefits from cluster assignments to keep the self-consistency between node representations and cluster assignments.•Extensive experimental results on three benchmark datasets have verified the effectiveness of the proposed method.

External IDs:dblp:journals/nn/ChenLZW25