It's not unreasonable that we might have a situation in which we are able to remove states from an automaton (for example, finding that f() dominates g() would allow us to change the sequence f();g() into just g() - if we call g then we must have called f previously, modulo some considerations).
The question is how much overhead does adding states to an automaton actually entail? It would be good to answer this both conceptually and empirically.
It's not unreasonable that we might have a situation in which we are able to remove states from an automaton (for example, finding that
f()dominatesg()would allow us to change the sequencef();g()into justg()- if we callgthen we must have calledfpreviously, modulo some considerations).The question is how much overhead does adding states to an automaton actually entail? It would be good to answer this both conceptually and empirically.