Quantitative CLTs in Deep Neural Networks
Date: 16 November 2023, Thursday
Time: 6:30 pm AEDT
Statistics
and Stochastic colloquium (part of the Colloquium Series of the
Department of Mathematics and Statistics) at La Trobe University jointly
organized with the Probability Victoria Seminar.
Contact the organizers: Andriy Olenko a.olenko@latrobe.edu.au, Kostya Borovkov kostya.borovkov@gmail.com
Speaker: Ivan Nourdin (Universitéit Lëtzebuerg, Grand Duchy of Luxembourg)
Abstract: In this talk, we
will study the distribution of a fully connected neural network with
random Gaussian weights and biases in which the hidden layer widths are
proportional to a large constant n. More precisely, we will explain how to prove quantitative bounds on normal approximations valid at large but finite n and any fixed network depth. This is based on a joint work with S. Favaro, B. Hanin, D. Marinucci and G. Peccati.
Zoom meeting link:
A
PDF file with the talk slides might become available for downloading
from https://probvic.wordpress.com/pvseminar/
prior to the talk
