Prototyping of machine learning (ML) solutions represents a pivotal stage in developing ML-enabled systems. In this course, the prototype serves as a means of communication and should demonstrate the technical feasibility and value to technical and non-technical stakeholders. But, in the context of the current ML solution prototyping process and tooling environment, non-technical stakeholders are limited in their ability to participate effectively, primarily due to the difficulty of understanding the specific ML solution strategy being implemented. In addition, valuable knowledge is lost during the prototype development process because the process is not sufficiently documented, preserved, or made easily accessible for future projects. To significantly improve the development of ML prototypes, we propose an extended ML prototyping process and tool support in the form of a toolbox. Preliminary implementations of some tools of the toolbox are presented.

When experimenting on solutions to Machine Learning problems, data scientists often integrate nonlinear workflows into Jupyter Notebooks to explore different approaches and evaluate the impact of changes such as using different Machine Learning models or adjusting parameters. This mode of working leads to Notebooks that are cluttered and difficult to navigate which complicates refining and reusing previous experiments later on. Jupyter Notebooks lack inherent support for such a mode of working. Therefore, we propose the JupyterLab extension HisTreethat provides an interactive tree-based representation of the experiment history in Jupyter Notebooks. Hereby, Note-book versions triggered by specified Notebook operations, are automatically saved and arranged in a tree structure. In this way, HisTreeallows data scientists to explore, compare, organize, and refine their past experiment approaches. In this paper, first, we introduce the concept of an experiment history tree model. This is followed by a comprehensive description of the functionality of HisTree,which aims to support data scientists in organizing experiments in Jupyter Notebooks. Initial feedback from user experiments shows that the tree-based experiment model is very promising and that the HisTreeextension is both useful and usable to conduct ML experiments in Jupyter Notebooks.

Continuous Delivery (CD) aims at reducing the cycle time from changes to software release while also increasing the software quality. To automate CD, delivery process models, defining all delivery activities need to be designed. Quality properties of delivery process models, such as maintainability, still oppose challenges. Previous research indicates that the quality of such models can be improved by aligning them with the software architecture. While software architecture knowledge is only incorporated implicitly, deep technical and process knowledge is required. On this basis, this paper introduces a new kind of delivery process models that focus mainly on software architecture knowledge. Hereby, we discard the current activity-centric view and shift to an artifact-centric view. Moreover, we outsource the required process- and technical knowledge to a transformation activities knowledge base. In order to make an artifact-based delivery process model executable, we provide a model-to-model transformation which constructs a CD pipeline from an artifact-based model with the help of the transformation activities knowledge base. We evaluated our approach by conducting a small industrial qualitative user study. It showed that low-experienced developers benefit from the reduced knowledge requirements of the artifact-based modeling approach.

The safety of software-based control systems plays an essential role in a vast number of applications. SynTACS is a tool that generates a framework for controller supervision, which enforces safety during runtime by utilizing the supervisory control theory of Ramadge and Wonham. In this paper, the results of a user study are presented in which it was investigated how far discrete-event systems, the underlying modeling formalism, are suitable to express safety requirements. Further, the usability of the tool was evaluated. In the second part, several concepts are introduced to support use cases that require real-time controllers due to unstable processes. Finally, two case studies are presented to show the applicability of both the tool and the new concepts.