ASIL: Augmented Structural Information Learning for Deep Graph Clustering in Hyperbolic Space

Graph clustering is a longstanding topic in machine learning. In recent years, deep learning methods have achieved encouraging results, but they still require predefined cluster numbers <inline-formula><tex-math notation="LaTeX">$K$</tex-math><alternatives><mml:math><mml:mi>K</mml:mi></mml:math><inline-graphic xlink:href="yu-ieq1-3661424.gif"/></alternatives></inline-formula>, and typically struggle with imbalanced graphs, especially in identifying minority clusters. The limitations motivate us to study a challenging yet practical problem: deep graph clustering without <inline-formula><tex-math notation="LaTeX">$K$</tex-math><alternatives><mml:math><mml:mi>K</mml:mi></mml:math><inline-graphic xlink:href="yu-ieq2-3661424.gif"/></alternatives></inline-formula> considering the imbalance in reality. We approach this problem from a fresh perspective of information theory (i.e., structural information). In the literature, structural information has rarely been touched in deep clustering, and the classic definition falls short in its discrete formulation, neglecting node attributes and exhibiting prohibitive complexity. In this paper, we first establish a <italic><underline>differentiable structural information</underline></italic>, generalizing the discrete formalism to continuous realm, so that we design a hyperbolic deep model (<monospace>LSEnet</monospace>) to learn the neural partitioning tree in the Lorentz model of hyperbolic space. Theoretically, we demonstrate its capability in clustering without requiring <inline-formula><tex-math notation="LaTeX">$K$</tex-math><alternatives><mml:math><mml:mi>K</mml:mi></mml:math><inline-graphic xlink:href="yu-ieq3-3661424.gif"/></alternatives></inline-formula> and identifying minority clusters in imbalanced graphs. Second, we refine hyperbolic representations of the partitioning tree, enhancing graph semantics, for better clustering. Contrastive learning for tree structures is non-trivial and costs quadratic complexity. Instead, we further advance our theory by discovering an interesting fact that structural entropy indeed bounds the tree contrastive loss. Finally, with an efficient reformulation, we approach graph clustering through a novel <italic><underline>augmented structural information learning</underline></italic> (<monospace>ASIL</monospace>), which offers a simple yet effective objective of augmented structural entropy to seamlessly integrates hyperbolic partitioning tree construction and contrastive learning. With a provable improvement in graph conductance, <monospace>ASIL</monospace> achieves effective debiased graph clustering in linear complexity with respect to the graph size. Extensive experiments show the <monospace>ASIL</monospace> outperforms 20 strong baselines by an average of <inline-formula><tex-math notation="LaTeX">$+12.42\%$</tex-math><alternatives><mml:math><mml:mrow><mml:mo>+</mml:mo><mml:mn>12</mml:mn><mml:mo>.</mml:mo><mml:mn>42</mml:mn><mml:mo>%</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href="yu-ieq4-3661424.gif"/></alternatives></inline-formula> in NMI on Citeseer dataset.