Research Center for Machine Learning#

The direct transfer of ML research to industry is the aim of the Fraunhofer Research Center for Machine Learning within the research cluster »Cognitive Internet Technologies«.

Research Center for Machine Learning

At the Fraunhofer Research Center for Machine Learning, the Fraunhofer Institutes IAIS, IOSB, ITWM and SCAI combine their many years of scientific expertise and know-how from the direct transfer of groundbreaking Machine Learning (ML) research into industry. It is one of three centers in the cluster Cognitive Internet Technologies (CCIT). At the Machine Learning Research Center, SCAI is working with the other institutes to explore innovative algorithms and methods that combine existing data-driven learning with automated modeling and domain-specific and application-specific expertise (Informed Machine Learning).

The Machine Learning Research Center faces the challenge of identifying and analyzing complex relationships and patterns in databases. In recent years, the availability of huge amounts of annotated data on the internet has led to significant advances in ML and artificial intelligence. However, industry-relevant data is often not freely available or widely annotated, so that companies are so far only able to use cognitive systems to a limited extent.

One goal of research at the Machine Learning Center is a new generation of reliable ML methods. These techniques systematically incorporate structural and procedural expertise into statistical training processes so that they provide robust and understandable results with minimal training data. Such Informed Machine Learning techniques have disruptive potential in extending the application and operation spectrum of ML.

The industry faces the challenge that learning outcomes, e.g. from modern deep networks, are often black-box systems whose decision-making is incomprehensible even to experts. However, transparency and traceability are indispensable in many economic applications, not only to estimate quality, reliability and risks, but also to make learning outcomes compatible with existing know-how and models.

This deficiency can be overcome by using the principle of the learned combination of modular building blocks. In Informed Machine Learning, these building blocks can consist of extensive expert knowledge, models and simulations..

Applications of Machine Lerning at Fraunhofer SCAI

Fraunhofer Research Center for Machine Learning