A tutorial on Federated Learning methodology for indoor localization with non-IID fingerprint databases

Abstract

This paper presents a tutorial on Deep Learning (DL) with Federated Learning (FL)-based indoor localization method for non-Independently and Identically Distributed (non-IID) fingerprinting databases. To this end, this paper explains systematic approaches for addressing privacy concerns and performance degradation issues in non-IID fingerprinting databases. The method presented in this tutorial entails the application of a personalized layer, model reliability, and Layer-wise local model’s Weight Change (LWC) information to FL. This tutorial provides intuitions to be considered by future researchers to improve the performance of FL-based fingerprinting localization by summarizing the above-mentioned methods into three FL-based techniques: high-complexity training for performance improvement of local training models, exact characteristics of the local model for global model aggregation, and Bayesian data fusion for probabilistic clustering, to improve FL-based indoor localization performance.

Keywords

• Federated Learning
• Fingerprinting
• Indoor localization
• Layer-wise local model Weight Change
• Non-IID database