http://bura.brunel.ac.uk/handle/2438/13037 Wed, 13 May 2026 19:43:27 GMT 2026-05-13T19:43:27Z http://bura.brunel.ac.uk/handle/2438/33268 Title: Reduced Precision Online Training for Spiking Neural Networks Authors: Kalganova, T; Fernandez-Hart, T; Knight, JC Abstract: ... Description: ... Thu, 01 Jan 2026 00:00:00 GMT http://bura.brunel.ac.uk/handle/2438/33268 2026-01-01T00:00:00Z http://bura.brunel.ac.uk/handle/2438/33180 Title: An energy-efficient dual-branch spiking neural network for epileptic seizure detection from electroencephalogram signals. Authors: Lou, X; Li, X; Meng, H; Hu, J; Xu, Y; Kong, H; Yang, J; Li, Z Abstract: Epileptic seizure detection from electroencephalogram (EEG) signals is critical for clinical diagnosis and long-term neurological monitoring. However, conventional artificial neural networks (ANNs) are often computationally expensive and energy demanding, which hinders their deployment in large-scale or real-time brain-signal analysis. Spiking neural networks (SNNs) provide a biologically inspired and energy-efficient alternative, yet existing architectures still struggle to balance accuracy and efficiency in EEG-based seizure detection. In this study, we propose an adaptive integrate-and-fire (AIF) spiking neuron model that dynamically adjusts its temporal behavior to capture diverse activation patterns. Based on this neuron, we develop a dual-branch spiking neural network (DBSNet), designed to decode multi-scale and multi-dimensional EEG features for improved seizure detection. We evaluate DBSNet on three public epileptic EEG datasets. Among SNN-based approaches, DBSNet consistently achieves state-of-the-art performance. On a large-scale dataset, it even surpasses the best-performing ANN while consuming only one-seventh of its theoretical energy, highlighting its efficiency advantage. These results demonstrate the potential of adaptive spiking architectures to achieve accurate and sustainable neural computing for EEG-based seizure detection, and they suggest a promising paradigm for broader applications in brain-signal processing. Description: Highlights: • Proposed AIF, a spiking neuron model with adaptive dynamics and high efficiency. • Developed DBSNet with SNN performance comparable to top-performing ANNs. • DBSNet consumes only one-seventh of the theoretical energy used by ANNs.; Data availability: https://github.com/xicheng105/DBSNet . Tue, 27 Jan 2026 00:00:00 GMT http://bura.brunel.ac.uk/handle/2438/33180 2026-01-27T00:00:00Z http://bura.brunel.ac.uk/handle/2438/33028 Title: Mitigating model coupling in semi-supervised segmentation via deep non-consistent mean teacher and fully collaborative learning Authors: Min, C; Lei, T; Wang, X; Wang, Y; Meng, H; Nandi, AK Abstract: Semi-supervised learning methods based on teacher-student frameworks have achieved remarkable success in image segmentation. However, popular teacher-student models are prone to early subnet coupling, which limits segmentation performance. Moreover, most existing approaches rely on strong-weak perturbation schemes for consistency learning, overlooking peer-level supervision between different perturbations and failing to fully exploit the potential information from unlabeled data. To address the above issues, we propose ComMatch, a novel semi-supervised image segmentation method built upon deep non-consistency and fully collaborative learning. Specifically, a deeply non-consistent mean-teacher structure is designed, which expands the learning space by constructing deep inconsistencies at both the data and network levels within a multi-stream learning framework and can effectively alleviate the problem of early subnet coupling. Meanwhile, to maximize the latent information from unlabeled data, a fully collaborative learning strategy is proposed, which explores the necessity of peer-level loss under deep inconsistency perturbations and further combines cross-level and peer-level losses to deeply mine the latent knowledge from unlabeled data. Experimental results show that the proposed ComMatch method surpasses the current state-of-the-art methods, achieving segmentation accuracies of 78.68% and 77.89% (1/16) respectively in the mIoU metric on the PASCAL VOC and Cityscapes datasets. Code is available at https://github.com/Minchongdan/ComMatch. Description: Data availability: Data will be made available on request. Thu, 12 Mar 2026 00:00:00 GMT http://bura.brunel.ac.uk/handle/2438/33028 2026-03-12T00:00:00Z http://bura.brunel.ac.uk/handle/2438/32908 Title: A Novel Modified Sine Cosine Algorithm for Reducing Side lobe Level of Linear Antenna Array Authors: Alturfi, AM; Salgotra, R; Almajidi, SD; Hussein, RA; Al-Raweshidy, H Abstract: ... Description: ... Thu, 01 Jan 2026 00:00:00 GMT http://bura.brunel.ac.uk/handle/2438/32908 2026-01-01T00:00:00Z