SemiPFL: Personalized Semi-Supervised Federated Learning Framework for Edge Intelligence

Recent advances in wearable devices and Internetof-Things (IoT) have led to massive growth sensor data generated edge devices. Labeling such for classification tasks has proven be challenging. In addition, by different users bear various personal attributes heterogeneity, rendering it impractical develop a global model that adapts well all users. Concerns over privacy communication costs also prohibit centralized accumulation training. We propose SemiPFL supports having no label or limited labeled datasets sizable amount of unlabeled is insufficient train well-performing model. this work, collaborate Hypernetwork the server, generating personalized autoencoders each user. After receiving updates from users, server produces set base models user, which locally aggregate them using their own dataset. comprehensively evaluate our proposed framework on public wide range application scenarios, health IoT, demonstrate outperforms state-of-art federated learning frameworks under same assumptions regarding user performance, network footprint, computational consumption. show solution performs without increasing performance increased number signifying effectiveness handling heterogeneity annotation. stability hardware resource three real-time scenarios.