Buongiorno a tutti,

Vorremmo segnalarvi che venerdì prossimo (21 Febbraio) in aula 1BC45 (Torre Archimede, Università di Padova) ci saranno due seminari per il ciclo di seminari in Probabilità e Finanza di:

Claudio Macci (Università di Roma Tor Vergata)
https://www.mat.uniroma2.it/~macci/
Title: Large and moderate deviations for Gaussian neural networks
Date: February 21, 2025, at 14:30, room 1BC45
Abstract: We prove large and moderate deviations for the output of Gaussian fully connected neural networks. The main achievements concern deep neural networks (i.e., when the model has more than one hidden layer) and hold for bounded and continuous pre-activation functions. However, for deep neural networks fed by a single input, we have results even if the pre-activation is ReLU. When the network is shallow (i.e., there is exactly one hidden layer) the large and moderate principles hold for quite general pre-activations and in an infinite-dimensional setting. Joint work with Barbara Pacchiarotti (Università di Roma Tor Vergata) and Giovanni Luca Torrisi (IAC-CNR).
Jonathan Tam (Università di Verona)
https://sites.google.com/view/jonathanyytam/home
Title: Extended mean-field control: a finite-dimensional numerical approximation
Date: February 21, 2025, at 15:30, room 1BC45
Abstract: We present a finite-dimensional numerical approximation for a class of extended mean field control problems. Our algorithm learns the value function on the whole Wasserstein domain, as opposed to a fixed initial condition. We leverage the approximation of the mean field problem by a finite-player cooperative optimization problem, due to the propagation of chaos, together with the usage of finite-dimensional solvers. This avoids the need to directly approximate functions on an infinite-dimensional domain, and allows for more efficient memory usage and faster computation times.

Vi aspettiamo numerosi!

Alberto Chiarini e Alekos Cecchin

Sito web del seminario: https://www.math.unipd.it/~chiarini/seminars/