Giovanni Pagliarini

I work on a few projects involving the application and formalization of new symbolic machine learning algorithms. The main research line revolves around the idea of equipping standard decision trees with the ability of performing temporal or spatial reasoning, so that they can natively handle time series, images, videos and, more generally, unstructured data that extends along temporal and/or spatial dimensions. This is done by adopting logical formalisms that are able to express entity-relation reasoning, for example, predicate logics or modal logics of space and time. The team I am working with is a mixture of logicians interested in modal logic, and machine learning people interested in interpretability; we have been following this lead for three years, achieving remarkable results for a few tasks of time series classification and image classification. Check out my publications.

My work is supervised by Prof. Guido Sciavicco under the ACLAI Lab initiative.