Answers are grounded in this paper and cite their sources.
A convenience copy of the Radoslaw manufacturing-company email network had silently discarded 98.2% of its message volume, three-quarters of its distinct directed edges and all directional information; we re-obtained the authentic 167-node network from the networkdata R package redistribution, which derives from KONECT, and validated it against the KONECT record. We re-implemented the greedy target-control algorithm of Gao and colleagues, validated it on five small test graphs, and quantified its approximation gap (3.6% on Radoslaw, 0.5% on the 1,005-node EU-core email network, which carries publicly available department labels for 42 organizational units). On both networks, controlling a locally clustered, structurally grown subset requires far fewer driver nodes than an equally sized random subset (Radoslaw: 5.2% versus 54.8% of the full-control budget at 50% coverage; EU-core: 2.6% versus 16.0%; means over 300 and 30 random draws). However, neither real departments (Wilcoxon p = 0.98) nor Louvain communities (p = 0.84–0.95, modularity 0.42) are more efficient to control than random subsets of the same size, and departments are significantly less efficient than structurally local clusters (p < 0.001). The advantage of local clustering is therefore specific to compact neighbourhoods grown by breadth-first search, not to community structure or organizational boundaries in general. On both networks, the full-control driver-node set has zero overlap with the top nodes by betweenness, in-degree, total degree, or PageRank; this follows largely from structure, since driver nodes have very low in-degree, consistent with the known tendency of driver nodes to avoid hubs. The analysis is static and cross-sectional; whether driver nodes change around reorganization events requires time-windowed data that remain to be acquired.
No takes yet. Share an insight, caveat, or question.
Spilios Dimakopoulos (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: