Senior researcher at IDSIA, member of the Imprecise Probability Group.
Research interests Probabilistic Graphical Models Machine Learning Bayesian Hypothesis Testing and Bayesian Hierarchical Models Imprecise Probability and Credal Classification Awards Best
Paper Prize at the International Environmental
Modelling and Software Conference 2002 (Lugano, Switzerland). Scientific events: PGM 2016 and ISIPTA 2017 I am co-chair and co-organizer of PGM 2016 , the Eighth International Conference on Probabilistic Graphical Models (Lugano, September 2016). The proceedings of PGM 2016 are published by JMLR.
I am co-chair and co-organizer of ISIPTA 2017 , the Tenth International Symposium on Imprecise Probability: Theories and Applications (Lugano, July 2017). The proceedings of ISIPTA 2017 are published by JMLR. Program committees IJCAI (2015, 2016): Int. Joint Conference on Artificial Intelligence UAI (2016): Conference on Uncertainty in Artificial IntelligencePGM (2014, 2016): European Workshop on Probabilistic Graphical ModelsECAI (2014, 2016): European Conference on Artificial IntelligenceISIPTA: International Symposium on Imprecise Probability: Theories and Applications: ( 2009, 2011, 2013, 2015, 2017)DMIN: International Conference on Data Mining (2011 , 2012 , 2013 , 2014 )TeachingApplied Statistics: Bach. of Management Engineering (SUPSI). Main topics: statistical inference and statistical process control. Uncertain Reasoning
and Data Mining: Master of Science in
Engineering (SUPSI). Main topics:
Bayesian networks and data mining. Co-teacher.
Bayesian hypothesis testing in machine learningSlides and code of our tutorial at ECML/PKDD 2016: G. Corani, A. Benavoli, J. DemsarComparing competing algorithms: Bayesian versus frequentist hypothesis testing.Naive Credal Classifier The open source
implementation of the naive credal classifier is the JNCC2.
Credal Model AveragingThe R implementation of credal model averaging for logistic regression is due to A. Mignatti. The algorithms are discussed in these two papers (link1, link2). The software and the marmot data set used in those papers is available here. |