Federated Learning in the Cloud for Analysis of Medical Images – Experience with Open Source Frameworks 

P., Jabłecki; F., Ślazyk; M., Malawski

Federated Learning (FL) is a novel technique that allows for performing the training of a global model without 0sharing data between entities. This research focused on the analysis of existing solutions for Federated Learning in the context of medical image classification. Selected frameworks: TensorFlow Federated, PySyft and Flower were tested and their usability was assessed. Additionally, experiments on classification of X-ray lung images with the use of the Flower framework were performed in a fully distributed setting using Google Cloud Platform.