18 dicembre 2015

Dott. Matteo BORROTTI - CNR-IMATI Milano
 

A Bayesian approach for multi-objective design of experiments

Abstract: Complexity arises in different fields of applications. The increasing number of variables and system responses used to describe an experimental problem limits the applicability of classical approaches from Design of Experiments (DOE) and Sequential Experimental Design (SED). In this situation, more effort should be put into developing methodological approaches for complex multi response experimental problems. In this work, we will develop a novel design technique based on the incorporation of the Pareto optimality concept into the Bayesian sequential design framework. One of the crucial aspects of the approach will involve the selection method of the next design points based on current information and the chosen system responses. The novel sequential approach has been tested on a simulated case study.

 

30 novembre 2015

Prof. Federico BASSETTI - Università di Pavia
 

Bayesian inference for panel AR(p) model with tree copula mixture innovations: an Application to Energy Data

Abstract: Motivated by the study of the energy market, we propose a Bayesian analysis for a panel AR(p) model where the multivariate distribution of the innovations is described by a mixture of “tree-copula” distributions. In this way we can assume that the innovations of the AR(p) are marginally normally distributed, without assuming their joint normality. The use of these copulas allows us to capture in a flexible way the dependence structure of the data. Tree copulas are particular types of Markov tree distributions in which the multivariate joint density is characterised through a suitable set of bivariate densities. We show that the fact that tree copulas rest on a simple graphical structure allows for efficient MCMC methods for posterior computations.

Joint work (in progress) with E. Nicolino, E. De Giuli, C. Tarantola

13 Novembre 2015

Prof. Roberto CASARIN

Bayesian Nonparametric Calibration and Combination of Predictive Distributions

Abstract:We introduce a Bayesian approach to predictive density calibration and combination that accounts for parameter uncertainty and model set incompleteness through the use of random calibration functionals and random combination weights. The proposed Bayesian nonparametric approach takes advantage of the flexibility of Dirichlet process mixtures, to achieve any continuous deformation of linearly combined predictive distributions. The weak posterior consistency of the Bayesian nonparametric calibration is provided under suitable conditions for unknown true density. We provide simulation examples with fat tails and multimodal densities, and applications to density forecasts of daily S&P returns and daily maximum wind speed at the Frankfurt airport.

Joint with F. Bassetti and F. Ravazzolo)

30 Ottobre 2015

Dott. Stefano PELUSO - Università Cattolica del Sacro Cuore

Parametric and Semiparametric Bayesian Financial Risk Models

Abstract We propose a Bayesian nonparametric model to estimate rating migration matrices and default probabilities using the reinforced urn processes (RUP) introduced in Muliere et al. (2000). The estimated default probability becomes our prior information in a parametric model for the prediction of the number of bankruptcies, with the only assumption of exchangeability within rating classes. The Polya urn construction of the transition matrix justifies a Beta distributed de Finetti measure. Dependence among the processes is introduced through the dependence among the default probabilities, with the Bivariate Beta Distribution proposed in Olkin and Liu (2003) and its multivariate generalization.

27 Maggio 2015

Prof. Mario PERUGGIA - Department of Statistics, The Ohio State University, Columbus, Ohio, USA

Reconciling two Popular Approaches for Summarizing Case Influence in Bayesian Models

Abstract Methods for summarizing case influence in Bayesian models take essentially two forms: (1) use common divergence measures for calculating distances between full-data posteriors and case-deleted posteriors, and (2) measure the impact of infinitesimal perturbations to the likelihood to gain information about local case influence.  Methods based on approach (1) lead naturally to considering the behavior of case-deletion importance sampling weights (the weights used to approximate samples from the case-deleted posterior using samples from the full posterior).  Methods based on approach (2) lead naturally to considering the curvature of the Kullback-Leibler divergence of the full posterior from the case-deleted posterior.  By examining the connections between the two approaches, we establish a rationale for employing low-dimensional summaries of case influence that are obtained entirely via the variance-covariance matrix of the log importance sampling weights.

This is joint work with Zachary Thomas and Steven MacEachern.

14 Maggio 2015

Prof. Mauro PREDA -  Università Cattolica del Sacro Cuore

Geographic Information Science. Teoria, metodi, strumenti e nuove prospettive di ricerca

Abstract Grazie alle reti, la crescente e diffusa disponibilità di prodotti software ed informazioni geografiche è in grado di offrire, pressoché a tutti, la possibilità di manipolare con facilità e rapidità i dati geografici, non solo, ma anche di creare una nuova base cartografica mondiale unica e partecipata v. l'esempio di OpenStreetMap In questo scenario, il “pensare geograficamente” parrebbe già parte integrante del normale processo di analisi, gestione  e soluzione strategica nelle Aziende a tutti i livelli ed in tutti i settori... ma è proprio vero?

Il workshop, partendo dalla definizione di  “informazione geografica”, con un approccio interdisciplinare di tipo filosofico, fisico, geografico, semantico, focalizzando, inoltre, l'attenzione sui modelli logici e fisici dei dati geografici, presenterà le basi concettuali ed un prototipo di "Sistema Informativo Geografico Esperto" con un esempio pratico applicato all'analisi e mappa del sentiment.

6 Maggio 2015

Dott. Stefano GLIOZZI - Senior Managing Consultant, IBM Italia

Junior Data Scientist in una azienda di consulenza e System Integration: considerazioni sugli ‘hard’ e ‘soft’ skill richiesti.

Abstract La domanda di Statistici, Data Miner, e Ricercatori Operativi (in breve ‘Data Scientist’) da parte delle aziende sta crescendo in modo tale da superare la offerta. Questo è un fenomeno globale ma se ne vedono i segni anche in Italia. Le Aziende di Consulenza e System Integration presentano una forte richiesta dei profili Junior di ‘Data Scientist’, per avviarle alla professione di Consulente di Management nel campo degli ‘Advanced Analytics’ (Business Intelligence, Modellazione Previsionale, Modellazione Decisionale). Nel corso degli anni più recenti ho avuto modo di assistere all’inserimento lavorativo di diversi giovani colleghi, di diverso curriculum studiorum, e ho notato alcune assenze di competenza, comuni a tutti i colleghi, rispetto ad alcuni skill (sia ‘hard’ che ‘soft’) utili per condurre con successo questa professione.  Il ‘Data Scientist’ che si occupa di Consulenza e System Integration, necessita infatti di alcuni skill specifici. Utilizzando il quadro metodologico del CRISP-DM, proverò a declinare, per ciascuna fase del processo di analisi dei dati, previsione e decisione, alcuni degli skill che non vengono spesso sufficientemente sviluppati durante la formazione universitaria, e che comunque l’aspirante ‘Data Scientist’ dovrebbe cercare di sviluppare durante la sua formazione.

12 Marzo 2015

Dott. Ciro CATTUTO - Scientific Director and Data Science Laboratory head, ISI Foundation

High resolution social networks: measuring, representing and modeling

Abstract. Smartphones and wearable devices are opening up a new window on human social behavior at the fine resolution of individual face-to-face interactions, impacting diverse research domains that span the social sciences, epidemiology, computer science, and more. In this talk we will focus on large-scale records of human face-to-face interactions measured by means of wearable sensors.