It is a great pleasure for the Nedo board to announce the following talk.

Juan Ignacio Perotti, formerly on board on the NETWORKS unit, is presenting the talk “Thermodynamics of the Minimum Description Length on Community Detection“, tomorrow, 2nd of October at 18 in classroom 1.

Abstract: In this talk I will introduce the Boltzmannian MDL (BMDL), a framework we have developed for statistical modeling [1]. Firstly, I will present the concept of universal codes and how they are related to different frameworks for statistical modeling. In particular, I will introduce the Refined MDL (RMDL), which is the “most effective” universal code, and which is the basis from where we develop ours [2]. Secondly, I will introduce the BMDL and how it should be used for model selection. Thirdly, I will reintroduce the BMLD; this time to address the statistical significance of the model selections. Fourthly and finally, I will show and example of its application on the problem of community detection in complex networks [3]. In particular, I will show how we can derive from the BMDL: i) Girvan-Newman modularity [4] and ii) Zhang-Moore method and criteria for community detection [5].

[1] Thermodynamics of the Minimum Description Length on Community Detection,
J.I. Perotti, C.J. Tessone, A. Clauset, G. Caldarelli (2018)
[2] The minimum description length principle,
P.D. Grünwald, MIT press (2007)
[3] Community detection in graphs,
S. Fortunato, Phys. Rep. 486, 75–174 (2010)
[4] Finding and evaluating community structure in networks,
M.E.J. Newman, M. Girvan, Phys Rev E 69, 026113 (2004)
[5] Scalable detection of statistically significant communities and hierarchies, using message passing for modularity,
P. Zhang and C. Moore, PNAS 111 (51) (2014)

BTW, Juan is the cousin of Diego Perotti, player of AS Roma.


Stay tuNED(O) for updates!