An Integer Programming Approach Reinforced by a Message-passing Procedure for Detecting Dense Attributed Subgraphs

Abstract

One of the recent challenging but vital tasks in graph theory and network analysis, especially when dealing with graphs equipped with a set of nodal attributes, is to discover subgraphs consisting of highly interacting nodes with respect to the number of edges and the attributes' similarities. This paper proposes an approach based on integer programming modeling and the graph neural network message-passing manner for efficiently extracting these subgraphs. The experiments illustrate the proposed method's privilege over some alternative algorithms known so far, utilizing several well-known instances.

Toward an Optimal Solution to the Network Partitioning Problem

“…Nevertheless, despite the widespread use of Modularity, it has been known to have certain limitations (see [12], [9] for more details). Notably, Modularity only takes into account the existing edges of the network, meaning it solely evaluates the goodness of a community based on its fit with the observed edges, while it fails to consider disconnected nodes (absent edges) within the same community.…”

Gain and Pain in Graph Partitioning: Finding Accurate Communities in Complex Networks

“…Therefore, maximizing Modularity aids in identifying high-quality communities within a network. However, despite the widespread use of Modularity, it has known limitations (refer to [22,34] for details). Notably, Modularity only accounts for existing edges in the network, evaluating the goodness of a community solely based on its fit with observed edges.…”