Machine Learning

Whether autonomous driving, developing new materials or analyzing and optimizing production processes (Industry 4.0) - the future cannot be imagined without artificial intelligence. Machine learning processes are among the key technologies in this field and they are becoming increasingly important for the intelligent analysis of ever-larger and more complex networked data sets. It’s about methods that generate knowledge from data. The systems are first of all “trained”, which enables them to learn from experience and improve themselves constantly.

Machine learning is a multifaceted, complex research field, in which the development and user communities merge seamlessly. The various learning methods can provide a preliminary sorting of the field, making it possible to distinguish between supervised learning, semi-supervised learning, unsupervised learning, reinforcement learning and active learning. Depending on the method of learning, different algorithms are employed. 

Using bibliometric procedures, Fraunhofer INT has examined this research field with regard to its main methodological approaches, the relevant players and their research areas. Bibliometry is an important quantitative analysis tool, and Fraunhofer INT uses it to be able to achieve a better assessment of a technology’s future importance and development progress.

Indicators such as the giant component provide information on the degree of cross-linking of individual research communities, and so indirectly give information on the potential scientific maturity level of a technology. Player analyses identify the key scientific protagonists, while overlay maps (see Fig. 1)  can project an institution’s key expertise fields on to a type of “map of the Sciences”, allowing the players to be aligned in relation to the landscape of research themes. 

The analyses confirm the general perception that machine learning in the factory of the future – in production, for example – will play an increasingly important role. The concept is highly developed and is becoming more and more widely used in a very broad variety of application fields. Within their technology and innovation management, companies should carefully analyze the future benefits they could reap from this technology, and arrange for timely planning accordingly.