A new approach to generalisation error of machine learning algorithms: Estimates and convergence

Michail Loulakis, Charalambos G. Makridakis

In this work we consider a model problem of deep neural learning, namely the learning of a given function when it is assumed that we have access to its point values on a finite set…