Skip to content

Performance Evaluation

Jump to bottom
Edi Muškardin edited this page Mar 23, 2021 · 3 revisions

All experiments were conducted on the Dell Lattitude 5410. PyPy 3.9 was used instead of CPython. With CPython, all experiments will take several times longer.

The runtime of the deterministic L* with respect to the size of the automaton (left) and alphabet size (right). In the first scenario, the alphabet size was set to 10, while in the second scenario number of states was set to 1000. We see a linear increase in runtime with respect to the automaton size. In the right graph, we see an interesting trend. From the alphabet size of 50 onward DFA's are less affected by the increase in alphabet size than Mealy and Moore machined. Riverst-Shapiro counterexample processing and no caching was used. Deterministic Evaluation

When inferring real-world systems, minimal overhead/computation introduced by AALpy pales in comparison to the interaction with the system. Therefore, when inferring deterministic systems use of caching is very much advised.

Overview

Home

Interactive Examples

Code Examples

Core Functionalities

Learning Black-box Systems

Conformance Checking

Setting Up Learning

Helper Functionalities

Random Automata Generation

Saving, Loading, Visualizing Automata

Advanced Use

Learning-based Testing

Learning-based Fuzzing

Model-Checking With PRISM

Other

Performance Evaluation

Clone this wiki locally