Rui Yang

Federated learning provides a natural framework for decentralized EEG analysis, but learning shared and transferable representations remains challenging because task-relevant neural patterns are often entangled with subject-dependent characteristics in local EEG data. Existing federated EEG methods mainly focus on backbone design, aggregation strategies, or personalized optimization, while representation-level control of what information is globally shared remains insufficiently explored. In this work, we propose Fed-VQCS, a vector-quantized content-style state-space disentanglement framework for robust federated EEG learning. Fed-VQCS introduces a plug-and-play Vector-Quantized Content-Style State-Space (VQCS) module that can be inserted after different EEG feature extractors. Given a latent EEG representation, a bidirectional state-space semantic extractor first produces a continuous task-semantic representation, which is further mapped to globally shared discrete content prototypes through vector quantization. A complementary residual pathway captures variation outside the continuous semantic estimate and is adversarially constrained to suppress task-discriminative information. During federated training, the shared model parameters and content codebook are aggregated globally, while the residual pathway is evaluated within each client’s local training pass. Experiments on BCI Competition IV-2a and PhysioNet, repeated over five random seeds, show consistent improvements in accuracy and Cohen’s κ across four EEG backbones. Comparisons with representative federated learning methods, sensitivity analyses, ablation studies, low-data adaptation, and representation probing further support the effectiveness and robustness of the proposed framework under heterogeneous subject distributions.

The advent of generative artificial intelligence (AI) technology has profoundly influenced the design field, thereby necessitating adaptations in contemporary design education. The benefits and limitations of employing generative AI in design emphasizes the significance of critical reflection in student learning. Therefore, in this paper, we propose a dual-phase design learning approach, integrating two learning phases (exploration and integration) with embedded reflective practice. To support the implementation of this learning approach, we developed a digital platform, Pocket Studio, offering students AI and lab resources for their learning activities. The proposed learning approach, applied in an undergraduate design course, received positive feedback from students. The results underscore the considerable potential of the dual-phase design learning approach in enhancing design education in the generative AI era and offer practical reference values for other modules in design disciplines to integrate AI into design education. This article was published open access under a CC BY licence: https://creativecommons.org/licenses/by/4.0/ .

Health monitoring and fault diagnosis of high-speed train traction systems are essential for maintaining reliable operation. To effectively process multisensor signals and avoid overfitting, ensemble learning methods are employed, leveraging multiple base models to integrate data from various sensors and enhance fault diagnosis performance. However, conventional ensemble frameworks are often burdened by excessive model parameters, limiting their applicability on edge computing processors. To address these challenges, this study proposes a novel dual attention-guided ensemble framework. This framework incorporates multiple multiscale feature attention (MFA) modules and a decision fusion attention (DFA) module, designed to capture critical features from multisensor signals, optimize the capacity of prominent feature extraction, and simultaneously reducing trainable parameters. The proposed ensemble framework is validated on the hardware-in-the-loop (HIL) simulation platform for high-speed train traction control systems, with experimental results demonstrating its superior effectiveness over several recently published ensemble learning methods.

The accurate determination of a material’s true stress-strain relationship is crucial for predicting its performance in engineering applications. However, traditional methods for determining material behavior often depend on extensive experiments. Although these methods are essential, they are inherently complex, time-consuming, and often suffer from limited accuracy and reliability when small sample sizes are involved. To address these challenges, this paper introduces a multi-fidelity deep neural network architecture that utilizes both low-fidelity and high-fidelity datasets to predict true stress-strain curves from small punch test data. The low-fidelity data are calculated using an empirical function based on the Swift model, providing an initial understanding of material behavior. In contrast, the finite element method provides high-fidelity data, enhancing the model’s ability to capture complex material characteristics. This framework merges low-fidelity and high-fidelity networks to estimate mechanical properties with low-fidelity and high-fidelity data. The effectiveness of the proposed approach has been demonstrated through experimental validation, highlighting its exceptional ability to generalize across material thicknesses and properties beyond the training set. The proposed approach outperformed traditional empirical methods and maintained strong performance even when applied to real-world materials.

Deep learning-based techniques have emerged as powerful tools for fault diagnosis. However, conventional deep learning methods require designing and training fault diagnosis models from scratch for each machine, necessitating a large-scale and high-quality labeled dataset. This limitation hinders the application of deep learning in practical settings, where labeled data is scarce and data distribution varies across different machines. To address these challenges, a novel self-supervised learning framework specifically designed for time-series data is proposed, aimed at enhancing cross-machine downstream fault diagnosis tasks with limited data. The proposed framework exploits inherent consistency between waveform and spectrogram representations, learning robust and transferable features from unlabeled data. Experimental results across three cross-machine fault diagnosis scenarios demonstrate that the proposed method outperforms existing state-of-the-art self-supervised methods, significantly reducing the reliance on labeled data and improving diagnostic performance.