Machine Learning Campus Minden: Ein Multi-Cloud Ansatz für Lehre, Konzepte und Erfahrungen

Within the project 'Machine Learning Campus Minden', a teaching concept was developed to further integrate the topic Machine Learning (ML) into the curriculum of the master's program in computer science at the Bielefeld University of Applied Sciences. To this end, we restructured specialized courses with ML context and created a new module 'Methods of Machine Learning'. Within this new course, all Machine Learning foundations, taught separately throughout different course modules, were centralized and built upon to allow a deeper dive into specialized topics for the more hands-on, advanced modules. These specialization modules teach students to understand demands, requirements and limitations within a real industry project on a big scale. Therefore, students explore scientific issues based on relevant data by doing group projects. If possible, cooperating companies provide the datasets. To be able to process large sets of data and fully utilize the discussed methods of Machine Learning, students are provided with cloud computing resources from three of the leading companies 'Microsoft Azure Machine Learning Studio' (MLS), 'Amazon Web Services' (AWS) and 'Google Cloud Platform' (GCP). This aggregation of cloud providers enables them to learn to integrate fitting cloud services to specific data science tasks successively. This paper discusses our experiences of creating the resulting teaching and researching concept and the lessons learned in integrating the multiple cloud service models for teaching services. It addresses both the technical and organizational needs and pitfalls and presents an evaluation regarding the cloud providers deployed in terms of domain-specific and practical problems.