jAlergia is minimal, modular, and efficient implementation of ALERGIA automaton learning algorithm in Java. It is developed without any dependencies and can be used either with your code, from command line, or from Python.
Learned models are saved in the .dot format and can be visualized with graphviz and used with AALpy. AALpy also includes python bindings to jAlergia .jar file.
jAlergia supports passive learning of
# Compile your own jar or download it from the repo
git clone https://github.com/emuskardin/jAlergia
gradlew jar
# gradlew.bat on Windows
java -jar alergia.jar -help
java -jar alergia.jar -input .\sampleFiles\mdpData1.txt -type mdp
# in case you run out of memory during FPTA construction
# - exten JVM memory with -Xmx, eg. -Xmx12g
# - and/or set -optim to 'mem'
class AlergiaExample {
public static void main() {
String path = "sampleFiles/mdpData2.txt";
double eps = 0.005;
ModelType type = ModelType.MDP;
String saveLocation = "jAlergiaModel";
OptimizeFor optimizeFor = OptimizeFor.ACCURACY;
List<List<String>> data = Parser.parseFile(path);
Alergia a = new Alergia();
a.runAlergia(data, type, eps, optimizeFor);
}
}Model learned by jAlergia can be loaded and visualized by AALpy, an open source active automata learning library written in Python.
from aalpy.learning_algs import run_JAlergia
from aalpy.utils import visualize_automaton
model = run_JAlergia(path_to_data_file='/sampleFiles/mdpData1.txt.txt', automaton_type='mdp', eps=0.005,
path_to_jAlergia_jar='alergia.jar')
visualize_automaton(model)Input files should follow the same syntax as defined in AALpy's Wiki. Following is the example of MDP input file. Note that numbers indicate inputs and capital character outputs.
C,0,A,1,A,0,B,1,C,1,A,1,C,0,D,0,C,0,A,1,C,1,A,1,C
C,1,B,0,C,1,A,0,B,0,C,0,D,0,C,1,A
C,0,A,1,A,1,A,0,B,0,C,1,A,1,C,0,D,0,C,1,B
C,1,B,1,D,1,C,0,D,0,C,0,A,1,C,1,C,0,D,1,C
...
Following the same logic, Markov chain input files would be sequences of inputs:
1,2,3,4,1,2,3,15,1,2,4
1,3,4,2,2,1,5,7,2,5
1,7,4,7,2,9,4,2,8,5
...
Stochastic Mealy machine input files follow the same syntax as MDP input file, just without the initial output.
jAlergia is a twin implementation of AALpy's Alergia implementation, and it is maintained as a part of the please cite: AALpy library. If you use jAlergia in the research context, please cite:
@inproceedings{aalpy,
title = {{AALpy}: An Active Automata Learning Library},
author = {Edi Mu\v{s}kardin and Bernhard K. Aichernig and Ingo Pill and Andrea Pferscher and Martin Tappler},
booktitle = {Automated Technology for Verification and Analysis - 19th International
Symposium, {ATVA} 2021, Gold Coast, Australia, October 18-22, 2021, Proceedings},
series = {Lecture Notes in Computer Science},
publisher = {Springer},
year = {2021},
}
If you have research suggestions, or you need specific help concerning your research, feel free to start a discussion or contact edi.muskardin@silicon-austria.com. I am happy to help you and consult you in applying automata learning in various domains.