Bisimulations over DLTS in O(m.log n)-time

Abstract

The well known Hopcroft's algorithm to minimize deterministic complete automata runs in $O(kn\log n)$-time, where $k$ is the size of the alphabet and $n$ the number of states. The main part of this algorithm corresponds to the computation of a coarsest bisimulation over a finite Deterministic Labelled Transition System (DLTS). By applying techniques we have developed in the case of simulations, we design a new algorithm which computes the coarsest bisimulation over a finite DLTS in $O(m\log n)$-time and $O(k+m+n)$-space, with $m$ the number of transitions. The underlying DLTS does not need to be complete and thus: $m\leq kn$. This new algorithm is much simpler than the two others found in the literature.