Participants learn the skills to develop reproducible workflows for data analysis and how to build there own tools to do so. An important learning aim is to develop a profound transfer knowledge that enables participants to answer questions such as the following ones: Why is reprodubility important in science? How can analytical workflows be designed to be as reproduible as possible? How can trustworthiness and applicability of machine learning models be assessed and quantified, especially since the reproducibility of training such models is difficult? Challenges, opportunities, limitiations and risks of the introduced methods are discussed. Understanding such intuitively is another important learning aim.
This course aims to deepen the participants‘ knowledge base and technical skills in the field of developing reproducible workflows to analyse scientific data and building software tools. Special focues lay on building models for pattern detection in Earth observation data using deep neural networks and machine learning, applying techniques to assess model trust and model applicability, implementing collaborative software development principals for automating development environments and utilizing machine-to-machine communication. The contents of the course are theoretically introduced, before they are practically applied and implemented using programming languages such as R or Python.