1 paper
Dennis Mouwen, Sicco Verwer, Azqa Nadeem
We present a method to learn automaton models that are more robust to input modifications. It iteratively aligns sequences to a learned model, modifies the sequences to their align…