Open Adda0 opened 1 year ago
I am currently working on it. It seems that the main inefficiency follows from the epsilon-closure computation. It seems that the best (performance-wise) way is to project the transition relation only to epsilons and then to use Purdom's or Floyd-Warshal's algorithm.
We need to optimize the algorithm for the removal of epsilon symbols from an automaton. To achieve that, we need to work on optimizing:
Mata::Nfa::remove_epsilon()
implementation.