Periodic Reporting for period 1 - AutoML (Unlocking the potential of machine learning for SMEs by automated machine learning)
Période du rapport: 2020-07-01 au 2021-12-31
Machine learning (ML) techniques that make predictions based on data are key to the modern data-driven industry. However, SMEs usually cannot afford the necessary specialist knowledge and therefore cannot benefit from ML. In this project, we evaluated whether our open-source prototype for automated machine learning (AutoML), Auto-sklearn, has the potential to lift this barrier.
Throughout the POC project we:
* analyzed the global market of existing AutoML solutions and to what extent these solutions are relevant to companies within Germany and the EU,
* studied the demand for AutoML solutions among data scientists and ML-practitioners and identified a strong interest in AutoML,
* researched how and for what our prototype is currently used and identified potential future extensions,
* integrated research findings from the ERC grant Beyond Blackbox into our open-source AutoML toolkit to further improve it (yielding the publication “Auto-sklearn 2.0: Hands-free AutoML via Meta-Learning” by Feurer et al. (2021)), and
* studied commercialization strategies for open-source and identified potential directions for our prototype.
To conclude, our analysis revealed several promising new directions for commercializing our AutoML technology which should be pursued further.
Throughout the POC project we:
* analyzed the global market of existing AutoML solutions and to what extent these solutions are relevant to companies within Germany and the EU,
* studied the demand for AutoML solutions among data scientists and ML-practitioners and identified a strong interest in AutoML,
* researched how and for what our prototype is currently used and identified potential future extensions,
* integrated research findings from the ERC grant Beyond Blackbox into our open-source AutoML toolkit to further improve it (yielding the publication “Auto-sklearn 2.0: Hands-free AutoML via Meta-Learning” by Feurer et al. (2021)), and
* studied commercialization strategies for open-source and identified potential directions for our prototype.
To conclude, our analysis revealed several promising new directions for commercializing our AutoML technology which should be pursued further.