The purpose of these 3 seminars, in the format close to a conference tutorial, is to offer doctoral students or researchers who wish to learn about artificial intelligence approaches for reasoning under partial knowledge, an introduction and an overview of existing approaches and tools, as well as a presentation of recent applications of this framework to different fields.
The first seminar will be more introductory and the two following seminars will develop, in a didactic way, recent research topics.
AI models for reasoning under partial knowledge: from theory to applications
Davide Petturiti
Department of Economics, University of Perugia (Italy).
email : davide.petturiti@unipg.it
https://sites.google.com/site/davidepetturiti
Biography
Davide Petturiti is Associate Professor in Mathematical Methods for Economics and Actuarial and Financial Sciences, in the Department of Economics, University of Perugia (Italy) since 2019. He obtained his teaching qualifications in Italy (ASN) in Mathematical Analysis, Probability and Statistics in 2021, and in Mathematical Methods for Economics and Actuarial and Financial Sciences in 2020. His research work focuses on artificial intelligence and decision theory. He is the author of more than 60 international publications including 26 articles in international journals and 23 book chapters.
Non-additive uncertainty measures and integrals emerged in AI and decision theory to overcome limitations of classical probability theory in addressing situations of partial knowledge. The term partial knowledge broadly refers to all cases where a complete probabilistic description of a problem is not available, like in presence of misspecification and ambiguity. This requires to deal with sets of probability measures or their envelopes, and paves the way to new classes of models that go beyond additivity. Nowadays, these theories consolidated in a vivid branch of AI that has several applications and poses the solution of challenging computational problems.
The goal of this series of seminars is to provide a panorama on the main tools for reasoning with partial knowledge, and present some recent applications in different domains. The seminars are structured as follows.
- Friday February 17, from 2 p.m. to 4 p.m.
Title: A glimpse of non-additive measures and integrals: theory and computational challenges
Abstract: In this seminar we present the main classes of models (lower/upper probabilities, belief functions, possibility measures, Choquet integrals and their generalizations), highlighting their connection to probability theory and expressive power. We also discuss interpretation and computational issues.
- Wednesday February 22, from 2 p.m. to 4 p.m.
Title: Applications in information fusion and inference: the statistical matching problem
Abstract: In this seminar we consider a specific integration problem, called statistical matching, referring to integration of data sets where some variables are separately observed and some others are observed in all the data sets. We face the problem in the framework of belief functions by computing inner/outer approximations, and showing advantages in making inference.
- Friday February 24, from 2 p.m. to 4 p.m.
Title: Applications in finance: pricing problems under ambiguity
Abstract: In this seminar we consider a pricing problem in a market with frictions in the form of bid-ask spreads. We first analyse a normative principle generalizing the standard no-arbitrage condition and derive a sound pricing rule. Next, we address the dynamic case by presenting some recent results on stochastic processes in the framework of belief functions.
Conference room, SCAI, Esclangon building, 1st floor
Pierre and Marie Curie Campus, Sorbonne University
4 place Jussieu, 75005 Paris