https://bura.brunel.ac.uk/handle/2438/8621 Fri, 09 Oct 2026 02:40:44 GMT 2026-10-09T02:40:44Z https://bura.brunel.ac.uk/handle/2438/33955 Title: Out-of-Distribution Detection Through Transformation Rectified Activation Authors: Li, Maozhen; Liang, Yu; Qi, Man; Liu, Guanjun Abstract: Rectified activation is a highly effective post-hoc method for out-of-distribution (OOD) detection, designed to enhance the robustness of Deep Neural Networks (DNNs). It operates by truncating high neuron activations to prevent overconfident predictions on unfamiliar data. However, this approach has two key limitations: it only truncates outlier activations without optimizing the post-truncation relationship between in-distribution (ID) and OOD data, and its performance is often unstable due to the manual selection of the rectification threshold. In this paper, we introduce Transformation Rectified Activation (Trans-ReAct), a novel method that addresses these issues. Trans-ReAct leverages two core innovations: 1) it applies mathematical transformations to strategically amplify the activation differentiation between ID and OOD data, and 2) it introduces a dynamic threshold anchor that automatically adapts the rectification threshold to the ID data, ensuring stable performance. Evaluated on three benchmark datasets, Trans-ReAct significantly outperforms state-of-the-art methods, reducing the false positive rate at 95% true positive rate (FPR95) by up to 9.76% and increasing the Area Under the ROC Curve (AUROC) by up to 2.04%. Fri, 19 Dec 2025 00:00:00 GMT https://bura.brunel.ac.uk/handle/2438/33955 2025-12-19T00:00:00Z https://bura.brunel.ac.uk/handle/2438/33954 Title: High-Resolution Automated Millimeter Wave Imaging System Authors: Hu, Shiwei; Hu, Shaoqing; Rao, Xin; Chen, Xiaodong; Nilavalan, Nila Abstract: Millimetre-wave imaging has promising applications in security screening, non-destructive testing and medical diagnostics due to its high resolution, non-ionising radiation and ability to penetrate optically opaque materials. However, the use of conventional high-resolution systems based on synthetic aperture radar (SAR) usually requires the use of mechanical scanning devices or tracking equipment for coherent aperture synthesis. In this paper, we propose a millimetre wave imaging system consisting of a FMCW (frequency modulated continuous wave) radar module, a programmable electromechanical XY scanning platform and dedicated control software. The system automatically acquires the echo data on a preset scanning grid, and then reconstructs the image by distance FFT and spatial frequency domain phase compensation. Experimental validation of the method for metal detection shows that the method can clearly distinguish target features and accurately locate the depth. Mon, 25 Aug 2025 00:00:00 GMT https://bura.brunel.ac.uk/handle/2438/33954 2025-08-25T00:00:00Z https://bura.brunel.ac.uk/handle/2438/33951 Title: AI-Driven Sparse MIMO Array Optimization and Deep Unfolding for Near-Field 3D Imaging Authors: He, Yuke; Hou, Shasha; Xu, Kuiwen; Qi, Xiaokang; Li, Xiaodan; Shen, Jianhua; Hu, Shaoqing; Li, Wenjun Abstract: ... Thu, 01 Jan 2026 00:00:00 GMT https://bura.brunel.ac.uk/handle/2438/33951 2026-01-01T00:00:00Z https://bura.brunel.ac.uk/handle/2438/33950 Title: MPHINE-KT: Heterogeneous Information Network Embedding With Metapath Knowledge Tracing Authors: Li, Meizi; Shao, Zhijie; Xu, Li; Li, Maozhen; Chen, Yunwen; Yang, Ru; Zhang, Bo Abstract: Contributions: This article proposes a new knowledge tracing (KT) model called heterogeneous information network embedding with metapath KT (MPHINE-KT) to model the knowledge state of learners and predict the accuracy of learners' future academic performance. To mine the transfer relationships between different knowledge points in sequences, the learning experience of learners is designed. The effects of individuality and commonality between learners' learning experience and abilities through heterogeneous information networks are analyzed, learners' individual attributes are introduced, and the problem of data sparsity is alleviated. The design improves the prediction effect of KT models on learners' future performance.Background: Online education platforms use KT models to analyze learners' online study. These models assess and predict learning effects, providing multidimensional feedback and guidance. They fail to adequately extract the features of the transfer relationships between different knowledge points for learners. Moreover, they lack discrimination regarding the differences in the characteristics of individual attributes of learners, and does not fully integrate the multiattribute features of learners, resulting in the sparsity of input data. This restricts the KT model's ability to predict learners' performance in answering future exercises.Intended Outcomes: The MPHINE-KT model analyzes the historical interaction sequence data of students and exercise problems, fully explores the relationships among learners, learning experience, and learning abilities, and more accurately predicts whether learners can answer new exercise problems correctly in the future, thus providing support for key educational aspects such as personalized education.Application Design: The MPHINE-KT model first designs a learner's learning experience recognition module based on a learner behavior graph for capturing learners' complex representations in the learning process, and then constructs a learner heterogeneous information network. Finally, as an additional input, the learner's feature vector is put into the gated recurrent unit (GRU) temporal prediction model, so as to improve the accuracy of predicting learners' future learning performance.Findings: Experiments are conducted on several real-world datasets, and the experimental results prove that the proposed model has a better performance compared to existing popular KT models. Mon, 15 Jun 2026 00:00:00 GMT https://bura.brunel.ac.uk/handle/2438/33950 2026-06-15T00:00:00Z