In this talk, I will present our work on active learning of timed languages recognizable by Event-Recording Automata (ERA). Active learning is a type of model inference approach to learn an unknown language by making membership and equivalence queries to a Teacher. In this work, we introduce an active learning algorithm, tLSep, for learning ERA-recognizable languages. Our framework employs a method known as grey-box learning, which enables the learning of ERA with the minimum number of states.
I will first introduce the active learning framework for inferring languages. I will then introduce the class of timed languages recognizable by ERA. Finally, I will describe our result on the learning of ERA-recognizable languages.
This talk is based on our ATVA 2024 paper with Sayan Mukherjee and Jean-François Raskin.
Short Bio:
Anirban Majumdar is currently visiting various research institutes in India. Earlier he was a post-doctoral researcher in the Verification group at Université Libre de Bruxelles, working with Jean-François Raskin. He did his PhD from ENS Paris-Saclay under the supervision of Patricia Bouyer-Decitre and Nathalie Bertrand. His research interests include synthesis of reactive systems, automata learning, distributed systems.
