I am currently a postdoctoral researcher at Imperial College in London. Before this I was a postdoc at the CWI (Centrum Wiskunde & Informatica) in Amsterdam. My research interests are in (large-scale) optimization and applied probability and their applications in machine learning. I am interested in applying these theoretical insights to contribute to explainable and human-centered AI, and the safe and ethical use of large datasets.
I obtained my PhD cum laude from the University of Bologna in Financial Mathematics as part of the Marie-Curie Industrial Doctorates and Horizon2020 project WakeUpCall under the supervision of Prof. Andrea Pascucci and Prof. Cornelis W. Oosterlee. I have obtained my Master's degree in Quantitative Finance at the VU Amsterdam and my Bachelor's degree in Applied Mathematics from the Delft University of Technology.
My current work:
Optimization: the development of stable and robust optimization algorithms
Neural networks: analytical expressions for the network output under simplifying assumptions, loss surface, the effects of the optimization method and its hyperparameters
Applications: the development of machine learning methods which are robust, human-centered, explainable and take into account privacy and ethical constraints; applications of machine learning in smart cities and finance
May 2019 "Generalization in fully-connected neural networks for time series forecasting" got accepted to the ICML Time Series Workshop
May 2019 "The effects of optimization on generalization in infinitely wide neural networks" got accepted to the ICML Workshop on Understanding and Improving Generalization in Deep Learning