2019/07/10 by Bernhard K. Aichernig, Roderick Bloem, Aichernig, Bernhard K. +16
Computer Science · Engineering · #FOS: Computer and information sciences #Fuel Cells and Related Materials #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning and Algorithms #Software Reliability and Analysis Research #Software Testing and Debugging Techniques
paper · pdf · doi:10.48550/arxiv.1907.04708
openalex publication_date 2019/07/10 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
Models play an essential role in the design process of cyber-physical\nsystems. They form the basis for simulation and analysis and help in\nidentifying design problems as early as possible. However, the construction of\nmodels that comprise physical and digital behavior is challenging. Therefore,\nthere is considerable interest in learning such hybrid behavior by means of\nmachine learning which requires sufficient and representative training data\ncovering the behavior of the physical system adequately. In this work, we\nexploit a combination of automata learning and model-based testing to generate\nsufficient training data fully automatically.\n Experimental results on a platooning scenario show that recurrent neural\nnetworks learned with this data achieved significantly better results compared\nto models learned from randomly generated data. In particular, the\nclassification error for crash detection is reduced by a factor of five and a\nsimilar F1-score is obtained with up to three orders of magnitude fewer\ntraining samples.\n