Enhancing Decentralized Federated Learning with User Feedback Loops: A Novel Approach for Personalized and Adaptive Learning in IoT Environments
Arpita Sarker, Alexander Jesser, Markus Speidel · 2024
The burgeoning field of the Internet of Things (IoT) demands innovative approaches to data privacy and user-centric services. This study introduces a novel cross-platform application engineered using the Flutter framework, which orchestrates federated learning (FL) and natural language processing (NLP) to facilitate personalized event discovery within IoT environments. The application features an NLP-based chatbot for user interaction and employs Dendrite, a second-generation Matrix homeserver, to manage decentralized communication. Central to the system's design is the stringent upholding of data privacy: user data, including browsing history and application usage patterns, is processed locally on user devices to construct profiles that inform personalized event suggestions. The system's deployment of end-to-end encrypted communication underscores its commitment to user privacy and security. This integration of FL and NLP showcases a significant leap forward in the realm of privacy-preserving, personalized applications, charting a new course for user engagement in IoT.