Francesco D'Angelo

Francesco D'Angelo is a Researcher and Doctoral Assistant at the École Polytechnique Fédérale de Lausanne (EPFL), affiliated with the School of Computer and Communication Sciences (IC) and the Institute of Computer Science and Communications Systems (IINFCOM). He is part of the Theory of Machine Learning Laboratory (TML) and enrolled in the Doctoral program in computer and communication sciences (EDIC) under the EPFL Doctoral School (EDOC). His research focuses on foundational aspects of machine learning, including sparse attention mechanisms, Bayesian neural networks, uncertainty estimation, and continual learning. He holds the position of Doctoral Assistant at the TML lab, where he contributes to advancing theoretical and practical applications of machine learning algorithms.

Research Interests: Francesco's work bridges theoretical computer science with applied machine learning. His key areas include developing robust models for uncertainty quantification, improving generalization in overparameterized networks through regularization techniques like weight decay, and exploring the interplay between transformer architectures and causal structure induction. He also investigates methods to mitigate catastrophic forgetting in continual learning scenarios and applies Bayesian approaches to enhance out-of-distribution detection capabilities.

Publications: His recent work explores topics such as sparse attention mechanisms in transformers, the necessity of weight decay in deep learning optimization, and Bayesian methods for uncertainty estimation. These contributions highlight his expertise in both foundational theory and practical algorithm design within the machine learning domain.

Labs & Teams: As part of the TML lab, he collaborates on interdisciplinary projects at the intersection of machine learning theory and applications, contributing to advancements in neural network architectures and probabilistic modeling techniques.