Dr. Daniel Horn Obtains Funding as an AI Trailblazer
- Artificial Intelligence
- Top News
- Research

AI models do not work at the push of a button; they have to be trained and adapted to the existing data each time they are used in new applications in order to deliver useful results. Currently, to find the best possible model parameter settings for each data set, computers typically test countless randomly generated settings simultaneously. "This requires very large computing power, which costs a lot of time, electricity and money," explains Dr. Daniel Horn. "If you could instead use an algorithm that specifically selects and tests only promising settings, it would be much more efficient and sustainable."
Algorithm to consider multiple criteria
This is exactly where the statistician will start with his hyperparameter tuning project and develop an algorithm that guarantees the best possible model quality. His goal is to optimize the respective model not only with regard to one criterion - for example, either the most error-free results possible or the fastest results possible - but to find a multi-criteria solution that reconciles these actually contradictory requirements.
Dr. Daniel Horn will be funded by the MKW for two years starting in June. In addition to the 175,000 euros from the state, the university will contribute ten percent of the funding as part of the "AI Trailblazer" program, bringing the total to around 200,000 euros. In the first selection round of the program, which started in 2020, Dr. Burim Ramosaj from the Department of Statistics had already obtained funding as an AI Trailblazer.
Contact for queries: