Dipartimento di Scienze
ATTENZIONE: Questa pagina contiene informazioni storiche
e non è più aggiornata.
link 15mo anniversario UdA Link esterno

RAPPORTI INTERNI
2005


Research Series


Title:Reti Neurali e Rischio di Credito: stato dell'arte e analisi sperimentale
Author(s):Giacomo di Tollo
Files: [pdf] [bib]
Tech.rep.number:R-2005-003
Abstract:Nel presente lavoro viene proposta l'applicazione delle reti neurali al problema della classificazione dell'insolvenza: abbiamo definito due architetture di rete e le abbiamo utilizzate per determinare se una azienda si rivelerà solvente o insolvente. Abbiamo utilizzato dati reali costituiti da un insieme di indicatori (di bilancio e dei modelli Centrale dei Rischi e andamentale) relativi a piccole e medie imprese italiane per un triennio. Dai risultati ottenuti emerge che le reti neurali rappresentano una buona soluzione al problema della determinazione del rischio di credito.

Title:Symmetry-breaking and local search: The case of LABS
Author(s):Andrea Roli
Files: [pdf] [bib]
Tech.rep.number:R-2005-002
Abstract:We report and discuss a series of experiments in which we compare the search space of LABS induced by modeling the problem with and without symmetry-breaking constraints. Furthermore, we compare the local search effectiveness in the two cases. In most of the instances analyzed, we observe that the total basin of attraction of global optima in the model with symmetry-breaking constraints is reduced by a factor that is higher than the search space reduction factor. We also experimentally find that local search is strongly affected by the size of global optima basins of attraction. To a certain extent, this behavior can explain why symmetry-breaking constraints have negative impact on local search.

Title:Modal logic and navigational XPath: an experimental comparison
Author(s):Massimo Franceschet and Enrico Zimuel
Files: [pdf] [bib]
Tech.rep.number:R-2005-001
Abstract:XPath is the core retrieval language of XQuery, the official query language for XML data. We empirically compare three query evaluation strategies for the navigational fragment of XPath known as Core XPath: a bottom-up algorithm based on model checking techniques for multi-modal logic, a first top-down procedure based on a technique to eliminate XPath filters, and a second top-down procedure that takes advantage of the pre/post plane representation of an XML tree. We implement the three methods and we benchmark the resulting XPath processors using a fragment of XPathMark, a recently proposed benchmark for XPath.