دانلود کتاب یادگیری فدرال – مروری جامع بر روش ها و کاربردها

Machine learning has made great strides over the past two decades and has been adopted in many application domains. Successful machine learning depends largely on access to quality data, both labeled and unlabeled. Concerns related to data privacy, security, and sovereignty have caused public and technical discussion on how to use data for machine learning purposes consistent with regulatory and stakeholder interests. These concerns and legislation have led to the realization that collecting training data in large central repositories may be at odds with maintaining privacy for data owners. While distributed learning or model fusion has been discussed since at least a decade, federated machine learning (FL) as a concept has been popularized by MacMahan and others since 2017. In the subsequent years, much research has been conducted – both, in academia and the industry – and, at the time of writing this book, the first viable commercial frameworks for federated learning are coming to the market.