Théorique, numérique
The purpose of this internship is to study via statistical physics methods the performances and properties of learning algorithms. An example is stochastic gradient descent which is the workhorse of optimization in high dimension and that is fundamental for the deep learning technology.