Abstract
Recommender systems have become ubiquitous in the past years. They solve thetyranny of choice problem faced by many users, and are utilized by many onlinebusinesses to drive engagement and sales. Besides other criticisms, likecreating filter bubbles within social networks, recommender systems are oftenreproved for collecting considerable amounts of personal data. However, topersonalize recommendations, personal information is fundamentally required. Arecent distributed learning scheme called federated learning has made itpossible to learn from personal user data without its central collection.Consequently, we present a recommender system for movie recommendations, whichprovides privacy and thus trustworthiness on multiple levels: First andforemost, it is trained using federated learning and thus, by its very nature,privacy-preserving, while still enabling users to benefit from global insights.Furthermore, a novel federated learning scheme, called FedQ, is employed, whichnot only addresses the problem of non-i.i.d.-ness and small local datasets, butalso prevents input data reconstruction attacks by aggregating client updatesearly. Finally, to reduce the communication overhead, compression is applied,which significantly compresses the exchanged neural network parametrizations toa fraction of their original size. We conjecture that this may also improvedata privacy through its lossy quantization stage.