Who is the best connected EC researcher? Centrality analysis of the complex network of authors in evolutionary computation

Co-authorship graphs (that is, the graph of authors linked by co-authorship of papers) are complex networks, which expresses the dynamics of a complex system. Only recently its study has started to draw interest from the EC community, the first paper dealing with it having been published two years ago. In this paper we will study the co-authorship network of EC at a microscopic level. Our objective is ascertaining which are the most relevant nodes (i.e. authors) in it. For this purpose, we examine several metrics defined in the complex-network literature, and analyze them both in isolation and combined within a Pareto-dominance approach. The result of our analysis indicates that there are some well-known researchers that appear systematically in top rankings. This also provides some hints on the social behavior of our community.

💡 Research Summary

The paper presents a comprehensive scientometric study of the co‑authorship network within the field of Evolutionary Computation (EC). Using the DBLP bibliography, the authors first identified a seed set of 3,773 researchers who had published in major EC venues (GECCO, PPSN, EuroGP, EvoCOP, and related workshops) over the preceding five years. By recursively gathering all co‑authors of these seed authors, they constructed an undirected graph comprising 7,712 unique authors and 13,169 papers.

Macro‑level analysis shows that the EC network exhibits classic small‑world properties: a high clustering coefficient (≈0.81), a relatively short average path length (≈10.9), a modest diameter (21), and a giant component that contains about two‑thirds of all nodes. These figures are comparable to, yet slightly more cohesive than, those reported for broader computer‑science collaboration networks.

The core contribution lies in the multi‑metric centrality assessment. Six standard centrality measures are computed for every node: (1) betweenness (the number of shortest paths passing through a node), (2) closeness (the reciprocal of the sum of shortest‑path distances), (3) degree‑based Bonacich power (a linear combination of a node’s degree and the power of its neighbours), (4) eigenvector centrality (the leading eigenvector of the adjacency matrix), (5) degree itself, and (6) a composite score derived from Pareto dominance.

Betweenness ranking highlights a small set of “information hubs”. The top ten include D.E. Goldberg, K. Deb, M. Schoenauer, H. de Garis, Z. Michalewicz, X. Yao, R.E. Smith, M. Tomassini, T. Bäck and K.A. De Jong. Their betweenness values follow a power‑law with an exponential cutoff, indicating a hierarchical hub structure.

Closeness ranking, which reflects how quickly a researcher can reach all others, yields a slightly different list: K. Deb, Z. Michalewicz, D.E. Goldberg, M. Schoenauer, B. Paechter, A.E. Eiben, D.B. Fogel, H.-G. Beyer, H.A. Abbass and M. Tomassini. The overlap with the betweenness list (five names) confirms that central actors tend also to be well‑positioned in terms of distance, yet the distribution is smoother, consistent with the network’s small‑world nature.

Bonacich power, emphasizing the combination of personal degree and the “power” of neighbours, brings D.E. Goldberg, M. Schoenauer and K. Deb to the top again, but also introduces D. Keymeulen, X. Yao, L.D. Whitley, T. Higuchi, H. de Garis and L. Kang. The appearance of Keymeulen, who is not a top betweenness or closeness actor, illustrates how power captures a different facet: the ability to leverage many well‑connected collaborators.

Eigenvector centrality, which assigns high scores to nodes linked to other high‑scoring nodes, lists D. Keymeulen, T. Higuchi, M. Iwata, I. Kajitani, N. Kajihara, M. Murakawa, E. Takahashi, H. Sakanashi, N. Otsu and M. Salami. These researchers form tightly knit clusters that dominate the spectral structure of the graph.

Pairwise scatter plots reveal that betweenness and closeness are moderately correlated, while power and eigenvector centralities are less so, confirming that each metric captures a distinct structural role. Consequently, relying on a single indicator would give a biased view of influence.

To integrate the complementary information, the authors adopt a Pareto‑dominance framework. Each author is represented as a six‑dimensional vector of centrality scores; an author A Pareto‑dominates B if A is at least as good on every metric and strictly better on at least one. The non‑dominated set (Pareto front) constitutes a multi‑objective “star” list, comprising researchers who perform well across all dimensions. This approach highlights scholars who are simultaneously bridges, rapid disseminators, and well‑connected power brokers.

The discussion interprets the findings sociologically. The EC community is shown to be organized around a relatively small elite of highly connected individuals who drive knowledge flow, foster collaborations, and shape research trends. The presence of well‑known figures (Goldberg, Deb, Schoenauer) across multiple rankings validates community intuition, while the emergence of less obvious names (Keymeulen, Higuchi) demonstrates the value of a nuanced, multi‑metric analysis.

Finally, the paper argues that the methodology—combining several centrality measures with Pareto analysis—offers a robust template for scientometric studies in other domains. It underscores the importance of multi‑objective evaluation when assessing scholarly impact, especially in fields where collaboration patterns are complex and evolving.