Optimisation, Dynamical Systems, and applications (Tuyen Trung Truong - University of Oslo, Norvège)

I will present some recent results on convergence in Gradient
Descent methods, which is basically the study of dynamics of gradient map, in
both finite and infinite dimensions. Possible applications are in Deep
Learning and finding solutions to PDE. A brief review about some major GD and
numerical methods (such as Newton's) is included. It will also be discussed
some open questions in Random Dynamical Systems which ultimately relates to
minima/ saddle points of maps.