Francesco Cagnetta

I am a Marie Skłodowska-Curie Fellow specialising in theoretical deep learning. My research focuses on the interaction between the layered architecture of deep learning models and the hierarchical structure of natural data, such as images and text. I am also interested in how this hierarchical structure influences data statistics and the effectiveness of learning algorithms.

I hold a PhD in nonequilibrium statistical mechanics from the University of Edinburgh, where I worked with Martin R. Evans and Davide Marenduzzo on developing theoretical models of the fluctuations of biological interfaces such as the cell membranes. I then joined Matthieu Wyart’s lab at EPFL, where we investigated the relationship between data structure and the sample complexity of deep learning methods.