Federated Learning of Explainable Artificial Intelligence (FED-XAI): A Review

Abstract

The arrival of a new wave of popularity in the field of Artificial Intelligence has again highlighted that this is a complex field, with issues to be solved and many approaches involving ethical, moral and even other issues concerning privacy, security or copyright. Some of these issues are being addressed by new approaches to Artificial Intelligence towards explainable and/or trusted AI and new distributed learning architectures such as Federated Learning. Explainable AI provides transparency and understanding in decision-making processes, which is essential to establish trust and acceptance of AI systems in different sectors. Furthermore, Federated Learning enables collaborative training of AI models without compromising data privacy, facilitating cooperation and advancement in sensitive environments. Through this study we aim to conduct a review of a new approach called FED-XAI that brings together explainable AI and Federated Learning and that has emerged as a new integrative approach to AI recently. Thanks to this review, it is concluded that the FED-XAI is a field with recent experimental results and that it is booming thanks to European projects, which are championing the use of this approach.