Adem Kikaj

This paper proposes NeSyAM, a novel neurosymbolic modeling paradigm for action model learning under partial observability, and presents a unified variational framework for theoretically analysing the limitations of existing methods compared to this proposed approach.