Interpretable graph clustering on massive attribute networks via multi-agent dynamic game

Abstract

The efficient clustering of attributed graphs is a critical and challenging problem that has attracted significant attention across various research fields. In this area, interpretation of the formation for attribute cluster configuration and balance between clustering quality and computational efficiency are two important issues. To solve these problems, in this paper, we model the attribute graph clustering problem as a multi-objective optimization problem and interpret the formation and evolution of cluster configuration by a new multi-agent dynamic game. By the effectively defining of feasible strategy mapping function and utility function for each node, the proposed framework demonstrates that the multi-objective optimization problem can be solved by calculating a series of coupled Nash equilibrium problems to achieve Pareto local optimal solution. Based on the theoretical analysis, we further propose a new attribute graph clustering algorithm. By updating the state variable to the convergence, one can automatically uncover rich semantic clusters on massive attribute graphs. The experiments on four public datasets show that the proposed method is effective and efficient.