Federated Machine Learning
Data-Governance in the Healthcare Industry
Machine learning has fundamentally changed image, text and speech recognition in recent years. Many of the methods applied in this area examine large amounts of centrally stored data. This may lead to risks in particular in the fields of data protection and data security. Centralizing data also concentrates liability risks, for example with regard to copyright infringements.
As a consequence, there are attempts to apply machine learning methods to various decentral data sets. This so-called federated learning (FL) offers new potentials to distribute risks and to gain new insights using less data. Analyses could, for example, be performed using the data sets of several market competitors. Particularly in the fast growing market for wearables used for therapy or diagnostics, solutions can be developed that offer prospects for both the users and the companies relying on the systems. However, this approach also raises new questions regarding competition law and especially antitrust law compliance.
At the ZLSR, Prof. Alfred Früh and his team are examining the potentials and risks of FL within the framework of a case study in cooperation with partner institutions and with a special focus on the healthcare industry in Basel.