<ul><li>Research on Graph Structure Learning (GSL) aims to improve graph-based clustering methods.</li><li>Current methods like GNNs and GATs struggle with sparse or noisy graph structures and may not fully capture underlying relationships between nodes.</li><li>The DeSE framework introduces Deep Structural Entropy to enhance graph clustering by quantifying structural information and using deep neural networks.</li><li>Extensive experiments show that DeSE outperforms eight unsupervised graph clustering baselines in terms of effectiveness and interpretability.</li></ul>

Unsupervised Graph Clustering with Deep Structural Entropy

Discover more