UniqueRank: Identifying Important and Difficult-to-Replace Nodes in Attributed Graphs

TL;DR

UniqueRank tackles the gap in node importance evaluation by jointly considering structural centrality and attribute-based uniqueness, addressing real-world replaceability. It combines a PageRank-style structural walk with an attribute-based walk in a two-graph Markov chain, controlled by damping $d$ and attribute-tradeoff $\alpha$, and reinforces the top candidates with a refinement step that maximizes dominance in both dimensions. Empirical results on terrorist, social, and supply-chain networks show that removing and attempting to replace UniqueRank-identified nodes yields greater efficiency reductions than conventional centrality or AttriRank baselines, with notable robustness to replacement criteria. The approach also extends to biomolecular domains, where atoms with unique environments and extreme forces are highlighted, suggesting broad applicability for identifying structurally critical yet hard-to-replace components across domains.

Abstract

Node-ranking methods that focus on structural importance are widely used in a variety of applications, from ranking webpages in search engines to identifying key molecules in biomolecular networks. In real social, supply chain, and terrorist networks, one definition of importance considers the impact on information flow or network productivity when a given node is removed. In practice, however, a nearby node may be able to replace another node upon removal, allowing the network to continue functioning as before. This replaceability is an aspect that existing ranking methods do not consider. To address this, we introduce UniqueRank, a Markov-Chain-based approach that captures attribute uniqueness in addition to structural importance, making top-ranked nodes harder to replace. We find that UniqueRank identifies important nodes with dissimilar attributes from its neighbors in simple symmetric networks with known ground truth. Further, on real terrorist, social, and supply chain networks, we demonstrate that removing and attempting to replace top UniqueRank nodes often yields larger efficiency reductions than removing and attempting to replace top nodes ranked by competing methods. Finally, we show UniqueRank's versatility by demonstrating its potential to identify structurally critical atoms with unique chemical environments in biomolecular structures.

Figures (14)

• Figure 1: Pipeline overview for evaluating network efficiency reduction after node removal and attempted replacement. The process involves identifying a key node, removing it, searching for a suitable replacement, and then measuring the resulting change in efficiency.
• Figure 2: A (synthetic) motivating example of an important application of UniqueRank: sub-graph of a terrorist network. Attribute categories are inspired by the Provisional Irish Republican Army network.
• Figure 3: Example plots for real networks, where the x-axis is structural importance (AttriRank) and the y-axis is log-scale attribute uniqueness (formally defined in Equation 7, as an inverse of similarity, defined in Equation 6 and consistent with hsu2017unsupervised, among a node and its neighbors).
• Figure 4: A conceptual view of UniqueRank’s Markov Chain, integrating two walks: one that captures the PageRank assumption (graph $G$) and one that captures the attribute assumption (graph $H$).
• Figure 5: An example illustration of the refinement step for selecting the top 2 nodes in a 10-node graph. The sampled nodes (connected to red lines) after the Markov Chain model step are used to identify a higher confidence set of nodes (yellow) during the refinement step.

Theorems & Definitions (4)

• Proposition 1
• proof
• Proposition 2
• proof