Distributed Estimation of Node Centrality with Application to Agreement Problems in Social Networks
Eduardo Montijano, Gabriele Oliva, Andrea Gasparri
Abstract
Eduardo Montijano, Gabriele Oliva, Andrea Gasparri
Abstract
Measures of node centrality that describe the importance of a node within a network are crucial for understanding the behavior of social networks and graphs. In this paper, we address the problem of distributed node centrality identification. In particular, we focus our attention on alpha-centrality, which can be seen as a generalization of eigenvector centrality, particularly suitable for graphs with asymmetric interactions. In this setting, our contribution is twofold: first we derive a distributed protocol where agents can locally compute their alpha-centrality index by means of local interactions; then we propose a novel consensus-algorithm running in parallel to the alpha-centrality estimator, which converges towards a weighted average of the initial conditions, where the weights are dictated by the alpha-centrality vector. The proposed algorithm finds application in social networks, where agreement protocols typically place more value on experts and influencers than on the rest of users. Simulations results are provided to corroborate the theoretical findings.
OpenAlex reports 4 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Measures of node centrality that describe the importance of a node within a network are crucial for understanding the behavior of social networks and graphs. In this paper, we address the problem of distributed node centrality identification. In particular, we focus our attention on alpha-centrality, which can be seen as a generalization of eigenvector centrality, particularly suitable for graphs with asymmetric interactions. In this setting, our contribution is twofold: first we derive a distributed protocol where agents can locally compute their alpha-centrality index by means of local interactions; then we propose a novel consensus-algorithm running in parallel to the alpha-centrality estimator, which converges towards a weighted average of the initial conditions, where the weights are dictated by the alpha-centrality vector. The proposed algorithm finds application in social networks, where agreement protocols typically place more value on experts and influencers than on the rest of users. Simulations results are provided to corroborate the theoretical findings.
Key concepts: Centrality, Katz centrality, Node (physics), Computer science, Network science, Social network (sociolinguistics), Estimator, Identification (biology)