Vai ai contenuti. | Spostati sulla navigazione | Spostati sulla ricerca | Vai al menu | Contatti | Accessibilità

logo del sistema bibliotecario dell'ateneo di padova

Brotto, Fabio (2017) Locating leading components in networks through the optimization of nonlinear modularity functions. [Magistrali biennali]

Full text disponibile come:

[img]
Preview
PDF
7Mb

Abstract

In this thesis we analyze the problem of bi-partitioning a given graph. We start from a natural linear view of the problem and then we move to a non linear extension that gives us a strict relation between the best bi-artitiong and the maximum of a continuous functions. Moreover we prove a new constrained formulation of the maximization problem with a convex and compact region as feasible set. From there we decide to approximate the objective function with a C1 function in order to use first-order local optimization algorithms. We show these algorithms, giving for each of them convergence theorems and related schemes. Finally we describe a probabilistic global method that we apply first on a toy model graph, given by a stochastic block model, and then on real world graphs. We compare results with some state of the art methods, trying to highlight what is the strength of our approach and which is the best algorithm to use.

Item Type:Magistrali biennali
Subjects:Area 01 - Scienze matematiche e informatiche > MAT/09 Ricerca operativa
Codice ID:56616
Relatore: Rinaldi, Francesco
Data della tesi:22 September 2017
Biblioteca:Polo di Scienze > Biblioteca di Matematica
Tipo di fruizione per il documento:on-line per i full-text
Tesi sperimentale (Si) o compilativa (No)?:Yes

Solo per lo Staff dell Archivio: Modifica questo record