Lots of data, little space:
Machine Learning with mini algorithms
The new ML2R Competence Center for Machine Learning (ML2R), officially opened at the end of January 2019, connects the different research institutions in this field in the Rhine-Ruhr metropolitan region, including the Fraunhofer IML. The institute's know-how is in demand regarding machine learning with restricted resources. Jens Leveling from Fraunhofer IML explains the project.
In modern, automated production facilities, large amounts of data are generated in addition to the actual products. Operating data of machines and industrial robots or measurement results of sensors are sources of such data. The evaluation of this data treasure trove holds enormous optimization potential: processes can be accelerated, quality can be improved, machines can be maintained preventively and robots can be employed more efficiently. Such applications often rely on machine learning methods for processing large amounts of data. Artificial neural networks in particular are able to learn from large sets of sample data and to develop models. These models then help production engineers, for example, to optimize manufacturing. However, such applications require very powerful computers. Especially for applications in Industry 4.0 and logistics, it is not always possible – for example due to a lack of bandwidth for data transmission – to send all information to powerful computers. Instead, the devices that record the data, such as sensors, have to take over part of the processing and analysis. Machine Learning with restricted resources now shall make it possible to reliably perform calculations using Machine Learning even on small devices such as smartphones or directly in sensors.
Resource restriction – in other words resource efficiency – has been an important research topic at Fraunhofer IML for many years. Driverless transport systems of the size of a transport or storage crate, both intelligent shelves or pallets can often only accommodate the smallest storage systems, batteries or sensors, which are therefore less powerful, not least due to legal requirements or standardized dimensions. Moreover, it is not always possible to send all information to a powerful central computer. Therefore, it is becoming more and more important to make better use of processor performance, memory or battery capacities.
This is why the Fraunhofer IML is developing solutions on making machine learning available on devices with limited computing power and memory as a part of its research center. One research approach is to simplify the ML algorithms so that they require less memory and computing capacity. Another course of action focuses on the development of hardware and software that is optimized for specific learning tasks. Respective demonstrators from the area of application of logistics will illustrate the research results.
The focus is on the human being
Today, machine learning forms the basis of a successful digital transformation of the economy. Intelligent systems can support human beings and learn from them. However, research has shown that many of the processes in learning systems are not really comprehensible for humans anymore. This is why Fraunhofer IML focuses on human beings when doing research on machine learning: On the one hand, they should be able to co-determine the learning processes of intelligent systems and, on the other hand, be able to continue to make their own decisions.
ABOUT THE AUTHOR
Jens Leveling is team leader “Data Driven Logistics” at Fraunhofer IML.
MORE ABOUT THE TOPIC
ML2R: Cutting-edge research on Machine Learning
The Competence Center Machine Learning Rhine-Ruhr aims at bringing Machine Learning (ML) technologies in Germany to a worldwide leading level. The ML2R Center is funded by the Federal Ministry of Education and Research (BMBF) as one of four nationwide ML nodes. The Technical University of Dortmund, the Fraunhofer Institutes for Intelligent Analysis and Information Systems IAIS in Sankt Augustin and for Material Flow and Logistics IML in Dortmund and the University of Bonn are involved. The common basis of the work in the ML2R Center is the research and development of modular ML applications.
In Modular Machine Learning, systems are built up from individual components and linked to complex architectures so that they can be used intuitively and are reusable in a flexible way. The researchers will focus on the following three main areas of research:
• Human-oriented Machine Learning focuses on humans and designs Machine Learning processes in such a way that decisions made with the help of Artificial Intelligence become understandable, traceable and validatable for human beings..
• Machine Learning with Restricted Resources makes it possible to reliably perform calculations using Machine Learning even on small devices such as smartphones or directly in sensors. This also includes learning on quantum computers, which circumvents the limitations of existing hardware.
• Machine Learning with Complex Knowledge integrates logical knowledge from various sources into learning systems to ensure reliable results even with small or insecure data sets.
Further information can be found here.