From 13th to 15th February 2019 we had our FeatureCloud Kick-off meeting. Within the next five years an awesome consortium will work on a ground-breaking novel federated cloud-AI infrastructure which only exchanges learned representations (the feature parameters theta θ, hence the name “feature cloud”) which are anonymous by default (privacy-by-design). Sensible real-world medical data remains secure at their place of origin. Collectively, the highly interdisciplinary consortium with expertise from AI and machine learning to medicine covers all aspects of the state-of-the-art machine learning value chain: assessment of cyber risks, legal and ethical considerations and international policies, development of state-of-the-art federated machine learning technology coupled to blockchaining, encompasing AI-ethics research and going towards explainable-AI. FeatureCloud’s goals are ground-breaking, challenging bold, obviously, but technically achievable. This is a pioneer project and can be seen as an European Apollo program to pave the way for a socially agreeable “big data era” for the benefit of future medicine.
We will help to contribute to the overall goal with three research streams: •1) WSI classification and alternative approaches •2) weakly-supervised and annotated interactive machine learning, fostering client-side learning with the human in the loop, •3) explainable AI (ex-AI) and explanation interfaces.
Explainable-AI shall foster re-traceability, thus transparency and trust to the approaches developed in the FeatureCloud project.
Raising legal, ethical, and social aspects make it mandatory to enable – on request – a human to understand and to explain why a machine decision has been made.
https://human-centered.ai/project/project-feature-cloud-federated-machine-learning/
