Description
The QNT algorithm currently initializes with a one-state PDFA and performs an equivalence query (EQ) to the teacher to obtain the first counterexample. However, a more intelligent approach would be to generate a tree by exhausting all known symbols from the lambda state. This approach could result in a better final model if we are time bounded, especially when the EQ process is slow, which is often the case in practice.
Proposed Changes
- Modify the initialization function to generate a tree that exhausts all known symbols before creating a PDFA and performing an EQ.
Description
The QNT algorithm currently initializes with a one-state PDFA and performs an equivalence query (EQ) to the teacher to obtain the first counterexample. However, a more intelligent approach would be to generate a tree by exhausting all known symbols from the lambda state. This approach could result in a better final model if we are time bounded, especially when the EQ process is slow, which is often the case in practice.
Proposed Changes