I am currently involved in the VeriLearn
project. Its aim is to apply formal verification methods to AI Systems in order to provide guarantees and justifications.
I am also interested in formal languages, and more specifically the study of tree or word transducers, including:
- The decidability of equivalence, origin-equivalence, and the potential existence of normal forms, for classes of tree transducers that extend the classical top-down definition, and their learnability (in the sense of symbolic learning).
- The extensions of those methods to wider formalisms on input or output, such as data trees, forests, DAGs...
(e.g. feature trees to model file systems for the CoLiS Project)