Direttore Vicario: Prof. Gabriele Grillo
Responsabile Gestionale: Dr.ssa Franca Di Censo


Selezionare una sezione
Parola da cercare

Prossimi Seminari

  • Nonintrusive reduced order models using physics informed neural networks
    Jan S. Hesthaven, Chair of Computational Mathematics and Simulation Science, EPFL, Lausanne, CH
    giovedì 29 ottobre 2020 alle ore 14:00 precise, Online seminar:

Seminari Passati

  • Bail-in vs bail-out: Bank resolution and liability structure
    Alessandro Sbuelz, Università Cattolica (Milano)
    martedì 18 febbraio 2020 alle ore 12:00, Aula Consiglio
    What is the joint impact of different resolution regimes and capital requirements on the optimal liability structure of a bank holding insured deposits and issuing non-bailinable debt and bail-inable Tier1-capital debt? We address this novel question and find that: 1) a credible bail-in resolution regime rules out extreme leverage and creates value by postponing default; 2) a positive probability of bail-out destroys credibility with dramatic effects on financial risk-taking, to the point of reversing the classical positive link between optimal leverage and growth prospects; and 3) a strict enforcement of the Basel III CET1 capital requirement strongly mitigates the impact of a non-credible resolution regime.
  • Kalman Filters for PDEs, a new hope
    Philppe Moireau, INRIA et Ecole Polytechnique, Paris, France
    lunedì 17 febbraio 2020 alle ore 16:00, Aula Saleri, VI piano
    Kalman filters for PDEs is an old topic which is theoretically seducing but leads to a so-called curse of dimensionality in its numerical implementation. In order to circumvent this curse of dimensionality, it is now classical to base the numerical strategies on reduced order modeling. This raises the question of the accuracy of the resulting estimator.
    Alternatives approaches based on reduced-basis decomposition of the covariance were also developed in the past years, but here with an additional question of stability of the resulting estimator.
    Using H-matrix based discretization strategies, a recent numerical tool developed for integral equations discretization, we here show how a full Kalman estimator can now be envisioned PDE models and their associated large discretization, hence freshening up the interest for this old theory.

  • Entropy-Transport distances between measures and metric measure spaces
    Nicolò De Ponti, Università degli Studi di Pavia
    martedì 11 febbraio 2020 alle ore 15:15, Aula seminari 3° piano
    After providing the necessary background material, we describe a class of distances coming from optimal Entropy-Transport problems, a recent generalization of optimal transport where also creation and destruction of mass is taken into account.
    Inspired by previous work of Gromov and Sturm, we then use these metrics to construct new meaningful distances between metric measure spaces with finite mass.
    This talk is based on a joint collaboration with Andrea Mondino and Giuseppe Savaré.
  • Semistatic and sparse variance-optimal hedging
    Paolo Di Tella, Technische Universitat - Dresden
    martedì 11 febbraio 2020 alle ore 10:30 precise, Aula Seminari MOX VI Piano
    We consider the problem of hedging a contingent claim with a “semistatic” strategy composed of a dynamic position in one asset and static (buy?and?hold) positions in other assets. We give general representations of the optimal strategy and the hedging error under the criterion of variance optimality and provide tractable formulas using Fourier integration in case of the Heston model. We also consider the problem of optimally selecting a sparse semistatic hedging strategy, i.e., a strategy that only uses a small subset of available hedging assets and discuss parallels to the variable?selection problem in linear regression. The methods developed are illustrated in an extended numerical example where we compute a sparse semistatic hedge for a variance swap using European options as static hedging assets. (Joint work with M. Haubold, M. Keller-Ressel)
  • A mathematical-physics approach to machine learning
    Pierluigi Contucci, Dipartimento di Matematica Università di Bologna
    giovedì 30 gennaio 2020 alle ore 14:00, Aula Saleri VI piano
    Artificial Intelligence is profoundly and quickly changing the technological profile of our society and yet machine learning, its disruptive spearhead, has almost no theoretical basis from a strictly scientific point of view. The talk will summarise the basic heuristic ideas on how it works together with a collection of open questions. A statistical mechanics framework will be used to formulate some of its problems on a mathematical-physics perspective and some preliminary results will be presented.

  • Going deep into shallowness
    Alessandro Verri, MaLGa - Università di Genova
    giovedì 23 gennaio 2020 alle ore 14:00, Aula Consiglio VII piano
    In the first part of my talk I quickly review the work we have been doing at UniGe in the last decades on machine learning. Ranging from theoretical to applied work I’ll highlight strengths and weaknesses of the regularization approach to learning. In the second part I argue, through examples and some surprisingly basic considerations, that while in certain application domains the widespread enthusiasm for deep learning is well justified, in others a more careful and critical approach might be the key to build truly intelligent systems.


  • Nonlinear Peridynamic Models
    Giuseppe Maria Coclite, Politecnico di Bari
    mercoledì 22 gennaio 2020 alle ore 15:15, Sala Consiglio 7° piano
    Some materials may naturally form discontinuities such as cracks as a result of scale effects and long range interactions. Peridynamic models such behavior introducing a new nonlocal framework for the basic equations of continuum mechanics. In this lecture we consider a nonlinear peridynamic model and discuss its well-posedness in suitable fractional Sobolev spaces.
    Those results were obtained in collaboration with S. Dipierro (Perth), F. Maddalena (Bari) and E. Valdinoci (Perth).
  • Stochastic Optimization with Multiple Time Scales
    Martin Glanzer,  University of Vienna
    martedì 21 gennaio 2020 alle ore 11:00 precise, Aula Seminari Terzo piano
    Real-world multistage stochastic optimization problems are often characterized by the fact that the decision maker may take actions only at specific points in time, even if relevant data can be observed much more frequently. In such a case there are not only multiple decision stages present but also several observation periods between consecutive decisions, where profits/costs occur contingent on the stochastic evolution of some uncertainty factors. We present a tailor-made modeling framework for such problems, which allows for a computationally efficient solution. We first establish new results related to the approximation of (Markovian) stochastic processes by scenario lattices. In a second step, we incorporate the multiscale feature by leveraging the theory of stochastic bridge processes. The ingredients to our proposed modeling framework are elaborated explicitly for various popular examples, including both diffusion and jump models. In particular, we present new results related to the simulation of compound Poisson bridges. Finally, we discuss a valuation problem of a thermal power plant, where implementing our multiscale modeling framework turned out to be particularly convenient. If time permits, we incorporate model ambiguity into the power plant valuation problem and show some numerical results.