Refine
Document Type
- Report (3)
Has Fulltext
- yes (3)
Is part of the Bibliography
- no (3)
Keywords
- MLOps (2)
- Machine Learning (2)
- E-learning (1)
- EXPLAIN-Project (1)
- Machine Learning Operations (1)
- Software Architecture (1)
Institute
- Fachbereich IV (3)
The University of Hildesheim owns many different software systems for teaching. So far,
no one has attempted to cover all systems and create an architectural description of e-learning at the University of Hildesheim. This report describes the applications relevant to education. These include software developed by the university, third-party tools provided by the university, cloud tools closely associated with the university, and tools provided by the university under contract. Implemented functionalities are explored,
and functional specifcations are described, including information about technical and operational requirements. Here, the software is primarily teaching tools. However, the report does not include software and development environments in which students are taught but rather tools that apply to teaching. The variety of tools in different courses makes it impossible to cover all the applications that are used by work groups or individuals.
The main goal of this investigation is to capture the university's e-learning state and describe the architecture - mainly to ease the development of applications for developers.
Further objectives are to identify new ideas and summarize current goals.
We conducted surveys of the teaching staff to determine the current status. Transcripts of interviews, tables provided by interviewed persons, and websites were used as references to prepare this document. We hope that future investigations will provide updates on current information and add information in this context.
MLOps have become an increasingly important topic in the deployment of machine learning in production. While Machine Learning Operations was predominantly used as a buzzword for methods in Machine Learning (ML) for the time being, since 2019, they are increasingly used in the context of deploying ML algorithms. This report is a protocol for a systematic literature review (SLR) that aims to determine the MLOps terminology and identify related activities. A further goal of the SLR is to identify where MLOps can be linked to classical software engineering. In addition, related automation techniques are considered.
The projected literature review aims to draw conclusions from papers that explicitly use the term MLOps or Machine Learning Operations with the objective to provide the necessary common baseline for future MLOps research and practice. This report thoroughly documents the SLR method, processes, and data material. We also gathered all relevant data to comprehend MLOps fully. Through our comprehensive analysis, we hope to provide valuable insights and recommendations for optimizing MLOps practices.
The EXPLAIN project (EXPLanatory interactive Artificial intelligence for INdustry) aims at enabling explainable Machine Learning in industry. MLOps (Machine Learning Operations) includes tools, practices, and processes for deploying ML (Machine Learning) in production. These will be extended by explainability methods as part of the project.
This study aims to determine to what extent MLOps is implemented by four project partner companies. Further, the study describes the ML use cases, MLOps software architecture, tools, and requirements in the companies perspective. Besides, requirements for a novel MLOps software architecture, including explainability methods, are collected.
As a result the interviews show that each of the interviewed industry partners use MLOps differently. Different tools and architectural patterns are used depending on the particular use case. Overall, most information we gathered focused on architecture decisions in the MLOps tool landscape used by the interviewed companies.
