Data Synthetization for Verification and Validation of Machine Learning Based Systems
Authors: Hamid Ebadi, Joakim Rosell, Thanh Bui, Lukáš Maršík, Jack Jensen, Martin Karsberg
Affiliations: Infotiv AB · RISE Research Institutes of Sweden · CAMEA · BERGE
Venue: DSC 2023 Europe VR, Antibes, France, 6–8 September 2023, pp. 1–5
Abstract
The incorporation of machine learning (ML) components presents challenges to the verification and validation (V&V) process due to the inherent opacity of ML systems. This paper introduces a tool-chain for generating synthetic datasets, aimed at facilitating search-based testing on a traffic monitoring system that utilizes machine learning.
The tool-chain is built around the BERGE graphical simulator and a scenario manipulator that generates synthetic data conforming to four key properties: relevant, complete, balanced, and accurate. The approach is demonstrated through an application on the CAMEA traffic control system, focusing on its ML-based License Plate (LP) detection component.
Keywords: Simulator, machine-learning, verification and validation, traffic scenario, synthetic data
Citation
Ebadi, H., Rosell, J., Bui, T., Maršík, L., Jensen, J., and Karsberg, M. (2023). Data synthetization for verification and validation of Machine Learning based systems. DSC 2023 Europe VR, Antibes, France, pp. 1–5.