Improving Model Inference of Black Box Components having Large Input Test Set
Résumé
The deterministic finite automata (DFA) learning algorithm L* has been extended to learn Mealy machine models which are more succinct for input/output (i/o) based systems. We propose an optimized learning algorithm L1 to infer Mealy models of software black box components. The L1 algorithm uses a modified observation table and avoids adding unnecessary elements to its columns and rows. The proposed improvements reduce the worst case time complexity. The L1 algorithm is compared with the existing Mealy inference algorithms and the experiments conducted on a comprehensive set confirm the gain.