Das SPA Team organisiert den neuen Track “Software Design Aspects for Industrial Machine Learning” auf der ISM 2023. Wir freuen uns auf SPA relevante Beiträge von anderen ForscherInnen und die gemeinsame Diskussion Vorort.

With the introduction of a new track to ISM 2023, we hope to further increase the community’s activities in SPA-related research areas. Our new track provides a platform for active participation through paper submissions and likely insightful discussions at the conference. We welcome your contributions to our track:

Software Design Aspects for Industrial Machine Learning
The increasing application of machine learning (ML) models in modern industrial settings raises more and more challenges along the lifecycle of a model: from training to deployment, execution, and adaptation. Due to the nature of fully automatized, just-in-time production lines, machine learning models not only need to be accurate, but also quick-to-serve predictions. Additionally, the long service life of industrial plants and associated high costs require dependable, trustworthy models that are easy to maintain. Changing parameters, e.g., production schedules, machine tooling, or material degradation, demand continuous monitoring of model performance and respective adaption. Such issues need to be addressed by the software components of a comprehensive industrial machine learning solution. In this track, we invite researchers to present software design aspects of new solutions to these challenges and lessons learned from real-world applications to avoid common pitfalls. For contributions to this track, topics of interest include, but are not limited to:

• Developments in ML pipelines, and workflows to address open issues in industrial application scenarios.
• Distributed processing for large-scale industrial machine learning applications (parallelization, containerization, orchestration).
• Application of established software engineering concepts for ML (CI/CD, unit testing).
• Validation of a model’s functional safety, prior to deployment (suitable execution environment, trustworthiness, interpretability).
• Software concepts for the interaction of simulation models and ML models in industry.