Research interests
I work at the intersection of deep learning / AI and computational neuroscience, bringing in tools from complex systems, information theory and statistics, to try to understand how natural and artificial systems process visual information. To do so, I employ both analytical methods and numerical simulations/optimization.
One of my main lines of work focuses on understanding how natural and artificial systems represent uncertainty. On the one hand, I train biologically inspired artificial neural networks to mimic inference processes in the brain. I apply these tools to the study of perception in neurotypical subjects and in autism spectrum disorders. On the other hand, I use this knowledge to improve uncertainty representations in deep learning models.
I also work in the field of algorithmic fairness, where the goal is to audit biases in AI systems in terms of protected attributes, understand the origin of these biases and how to mitigate them. We focus on ML systems for medical images and biomedical signals.
Sampling-based representation of uncertainty in the context of Bayesian inference.
