Pervasive computing promotes the integration of connected electronic devices in our living spaces in order to assist us through appropriate services. Two major developments have gained significant momentum recently: a better use of fog resources and the use of AI techniques. Specifically, interest in machine learning approaches for engineering applications has increased rapidly. This paradigm seems to fit the pervasive environment well. However, federated learning has been applied so far to specific services and remains largely conceptual. It needs to be tested extensively on pervasive services partially located in the fog. In this paper, we present experiments performed in the domain of Human Activity Recognition on smartphones in order to evaluate existing algorithms.

Pervasive computing promotes the installation of connected devices in our living spaces in order to provide services. Two major developments have gained significant momentum recently: an advanced use of edge resources and the integration of machine learning techniques for engineering applications. This evolution raises major challenges, in particular related to the appropriate distribution of computing elements along an edgeto-cloud continuum. About this, Federated Learning has been recently proposed for distributed model training in the edge. The principle of this approach is to aggregate models learned on distributed clients in order to obtain a new, more general model. The resulting model is then redistributed to clients for further training. To date, the most popular federated learning algorithm uses coordinate-wise averaging of the model parameters for aggregation. However, it has been shown that this method is not adapted in heterogeneous environments where data is not identically and independently distributed (non-iid). This corresponds directly to some pervasive computing scenarios where heterogeneity of devices and users challenges machine learning with the double objective of generalization and personalization. In this paper, we propose a novel aggregation algorithm, termed FedDist, which is able to modify its model architecture (here, deep neural network) by identifying dissimilarities between specific neurons amongst the clients. This permits to account for clients' specificity without impairing generalization. Furthermore, we define a complete method to evaluate federated learning in a realistic way taking generalization and personalization into account.Using this method, FedDist is extensively tested and compared with three state-of-the-art federated learning algorithms on the pervasive domain of Human Activity Recognition with smartphones.

No abstract