- University of Hildesheim ›
- Mathematics, Natural Sciences, Economics and Computer Science ›
- Institutes ›
- Institute of Computer Science ›
- Abteilungen ›
- Data Science (DS) ›
- Current Research Projects
Current Research Projects
Building models from experimental data: machine learning and model evaluation under dependencies and distribution shifts
Funding: German Research Foundation (DFG)
The analysis of experimentally obtained data is one pillar of knowledge generation in the natural sciences. It enables the formation of models describing natural phenomena; the ability of the obtained models to predict the behavior of observed systems is often evaluated on the basis of experimental data. However, widespread methodological tools for model building and model evaluation are based on independence and distribution assumptions that may violate experimentally obtained data in various ways. For example, the distributional properties of experimental data are determined by the choice of observational protocols and experimental parameters; the acquired data provide a representation of reality that is shaped by the process of measurement and the properties of available sensors. The goal of the project is to develop machine learning methods for model building and model evaluation that explicitly represent the experimental observation process and the resulting differences between nature and experiment. To this end, we develop approaches to appropriately correct for dependencies and distributional shifts in experimental data; we investigate the formal properties of the resulting procedures and the complexity of the optimization problems to be solved. Results of the project should reduce the experimental cost of model building and make conclusions based on experimental data more robust. In collaboration with research groups from cognitive psychology and geophysics, we aim to achieve advances on exemplary model building problems in the sciences (link).
Artificial intelligence for the detection of respiration in dairy cows (KAMI)
Funding: Bundesinstitut für Ernährung und Landwirtschaft (BMEL)
Using AI technologies and unlocking their potential opens up new opportunities to sustainably intensify milk production and use resources efficiently and competitively. AI tools and data-based applications enable a better understanding of complex interactions between the husbandry environment and the animals living in it. Animal welfare and health are improved, performance potentials are exploited and the use of medication is reduced. The project "Artificial Intelligence for Detecting Respiration in Dairy Cows (KAMI)" is developing a prototype for automated, computer vision based detection of respiration in cows. Respiration rate is one of the most sensitive parameters for vitality and health and indicates deviations from healthy state at an early stage. Until now, respiration can neither be recorded individually for an animal without contact-based sensors, nor can it be evaluated and translated into recommendations for better animal management.
Sustainable intensification of food production through resilient farming systems in West & North Africa (SustInAfrica)
Fördermittelgeber: European Union, Horizon 2020-Programme
SustInAfrica is a research project empowering West and North African smallholder farmers and small- and medium-sized enterprises (SMEs) to facilitate sustainable intensification of African farming systems. We aim to develop and deploy a reference framework on best agricultural practices and technologies, based on a systems approach, and successfully verified for their efficacy to intensify primary production in a self-sufficient, sustainable and resilient manner.
Within the project, our group specifically works on smart farming technologies for plant production (WP2). Here, the overall objective is to conceptualise and develop technical tools and solutions to monitor and manage plant performance for sustainable plant production intensification in Africa. In a general sense, we aim to develop a concept for promoting smart, open, and affordable monitoring technologies opting to improve plant production, plant health, water management, and the delivery of ecosystem services (link).