Interactive Machine Learning for Industrial Data Science
The overall goal of INTERACTIVE is to design and develop workflows and algorithmic methods that enable interactive machine learning in distributed edge computing environments. Resulting workflows and algorithms are implemented for two distinct Proof-of-Concept use cases: product quality prediction in high-performance automation, and predictive maintenance for cooling facilities.
The expected results and innovations are:
- Machine learning workflows and lifecycles designed for distributed AI architectures.
- Software libraries tackling the missing ground-truth problem by supporting incremental data labelling and model training.
- Federated learning approaches enabling model training and updates on edge devices without transferring large amounts of raw data to central servers.
- PoCs for both use cases allowing assessment of benefits on the technical, operational, and business level.
Partners: SIEMENS, STIWA, Hauser
Facts:
- Projektbeginn: April 2021
- Projektdauer: 24 Monate
- Budget: ca. 700k EUR
- Förderung: FFG
