Semi-supervised federated learning (SSFL) aims to leverage the vast amount of unlabeled data distributed across clients and a limited amount of labeled data held by the central server. However, SSFL faces a tough challenge of catastrophic forgetting, caused by discrepancies between local and global data distributions in non-independent and identically distributed (non-IID) settings, and exacerbated by noisy pseudo-labels. To deal with this problem, existing methods typically focus on mitigating the adverse effects of distribution divergence and refining the pseudo-labels. Differently, in this paper we tackle this problem from a data perspective by reducing the divergence between local and global distributions. Specifically, we propose global knowledge generative replay, which generates synthetic samples to complement the missing global knowledge during local training. Additionally, we introduce forgetting-aware model aggregation, a method that adaptively re-weights local models based on their degree of knowledge forgetting, resulting in a more robust global model. We conduct extensive experiments on widely-used benchmark datasets, and experimental results show that our method achieves state-of-the-art performance across various data settings, validating its effectiveness and superiority. The code will be available at https://github.com/lhq12/SemiFed.

In wireless power transfer (WPT) systems, the design of coupling coils plays a crucial role in power transfer. To increase the coupling coefficient, ferrite is added to conventional couplers. However, this can drastically reduce the power density of the system. In this paper, without changing the area of the transmitting coil, increase the power density of the system by increasing the number of the transmitting coil turns, combined with the application of ultra-thin magnetic material on the surface of the coupling coil. Based on the circuit principle, Model the equivalent circuit model of the system Analyze the influencing factors of the output power of the system. Then the 3D model was built by Ansys Maxwell to simulate the performance metrics under various scenarios, especially the coupling coefficient k and mutual inductance. Finally, it was verified through experiments that the combination of appropriately increasing the number of coil turns and adding ultra-thin magnetic materials will significantly increase the self-inductance, mutual inductance and coupling coefficient of the coupling coil, and improve the power density of the system.

With the wide application of face recognition systems, there is rising concern that original face images could be exposed to malicious intents and consequently cause personal privacy breaches. This paper presents DuetFace, a novel privacy-preserving face recognition method that employs collaborative inference in the frequency domain. Starting from a counterintuitive discovery that face recognition can achieve surprisingly good performance with only visually indistinguishable high-frequency channels, this method designs a credible split of frequency channels by their cruciality for visualization and operates the server-side model on non-crucial channels. However, the model degrades in its attention to facial features due to the missing visual information. To compensate, the method introduces a plug-in interactive block to allow attention transfer from the client-side by producing a feature mask. The mask is further refined by deriving and overlaying a facial region of interest (ROI). Extensive experiments on multiple datasets validate the effectiveness of the proposed method in protecting face images from undesired visual inspection, reconstruction, and identification while maintaining high task availability and performance. Results show that the proposed method achieves a comparable recognition accuracy and computation cost to the unprotected ArcFace and outperforms the state-of-the-art privacy-preserving methods. The source code is available at https://github.com/Tencent/TFace/tree/master/recognition/tasks/duetface.

Most of the approaches in the application of network coding either require the overhearing opportunity or have bad interaction with TCP. A new mechanism, named TCP-I2NC, is proposed in this paper to incorporate network coding into TCP in interference-free multi-radio multi-channel wireless mesh networks where there are nearly no overhearing opportunities. Multiple TCP flows are coded together and forwarded in block by hop-by-hop ACK and retransmissions in TCP-I2NC. Several encoding blocks are working simultaneously to effectively utilize available bandwidth. However the maximum of encoding blocks is adaptively adjusted according to the bandwidth, propagation delay and packet loss rate of a wireless link, and also back pressure algorithm is used to perform flow and scheduling control. The end-to-end delay is optimized so that TCP-I2NC is applicable to delay-sensitive applications. Simulations show that our mechanism both significantly improves the throughput of TCP and derives a relatively short end-to-end delay as losses increase. And the delay jitter of TCP-I2NC is also very small. TCP-I2NC also achieves complete fairness in resource allocation.