Agent Marketplaces and Deep learning in Enterprises: The COMPOSITION project by Dario Bonino and Paolo Vergori for the 41st IEEE Computer Society Signature Conference on Computers, Software and Applications on 4-8th July 2017 in Turin, Italy.

Abstract: The Factory of the Future vision focuses on a new generation of enterprises that will exploit latest breakthroughs in data sciences and in Internet-connected devices to effectively address the ever increasing pace at which data and goods are nowadays traded. Among the many challenges emerging in this domain, particularly important is the ability to enhance the connection between the value-production chains of factories and the outer network of supply partners, and service providers that enable enterprise business. In this paper, the authors highlight how the critical issue of connecting inner and outer value chains in factories can be effectively tackled through state-of-the-art IT solutions based on Machine Learning, Artificial Intelligence and Internet of Things. In particular, the approach to these challenges pursued in the COMPOSITION project is depicted and envisioned solutions are discussed.

Industry 4.0: Mining Physical Defects in Production of Surface-Mount Devices by Farshid Tavakolizadeh, José Ángel Carvajal Soto, Dávid Gyulai, and Christian Beecks for the 17th Industrial Conference on Data Mining (ICDM), 12-16 July 2017 in New York, USA.

Abstract: With the advent of Industry 4.0, production processes have been endowed with intelligent cyber-physical systems generating massive amounts of streaming sensor data. Internet of Things technologies have enabled capturing, managing, and processing production data at a large scale in order to utilize this data as an asset for the optimization of production processes. In this work, we focus on the automatic detection of physical defects in the production of surface-mount devices. We show how to build a classification model based on random forests that efficiently detects defect products with a high degree of precision. In fact, the results of our preliminary experimental analysis indicate that our approach is able to correctly determine defects in a simulated production environment of surface-mount devices with a MCC score of 0.96. We investigate the feasibility of utilizing this approach in realistic settings. We believe that our approach will help to advance the production of surface-mount devices.