Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Adam Ahmed Qaid Mohammed has a total of 63 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    UGSA-Net: Uncertainty-Gated Attention and Self-Supervised Alignment for Calibration-Aware Hyperspectral Image Classification

    by Houari Youcef Moudjib, Haibin Duan, Zain Anwar Ali, Adam A. Q. Mohammed

    Uniform spatial weighting in patch-based hyperspectral classifiers indiscriminately aggregates spectral information, even when individual pixels are already class-diagnostic, degrading accuracy and interpretability. We propose UGSA-Net, an ambiguity-guided spatial–spectral transformer that infers a per-pixel ambiguity field without direct label supervision. A lightweight ambiguity head maps each token to a scalar in (0,1), trained end-to-end with no target of its own, so that the field is shaped only by the training losses that reach it through the token mixer. This field dually modulates the token mixer: it biases query and key projections with a spectral angle mapper (SAM) prior to emphasise spectrally similar patches, and rescales token outputs to amplify contributions from ambiguous locations. The value pathway remains unmodulated, so the network reweights spatial information without distorting the spectral content entering the mixture. The ambiguity head and the classification backbone are trained end-to-end, following a self-supervised latent-alignment stage that uses no labels; model selection jointly scores overall accuracy (OA), few-shot accuracy and expected calibration error (ECE) on validation data. Across six benchmark scenes from four sensors, under a stratified pixel-level split in which training and test windows overlap, UGSA-Net exceeds 98.9% OA on every scene over three runs, and on Indian Pines it is comparable to competing architectures retrained under the same protocol. These values are upper bounds rather than estimates for new scenes: on two of the six scenes, almost no labelled pixel has a competing class inside the 9×9 input window. A single validation-tuned temperature reduces ECE from 0.24–0.28 to below 0.016 without changing any prediction, revealing that residual miscalibration is governed by the decision head rather than by the data representations. The benefit of ambiguity gating emerges when labels are scarce: with a reduced training set on Indian Pines, gating raised OA over the ungated model on all three partitions, in either direction, whereas, with the full training set, component effects fall within run-to-run variation.

  • Open Access

    Energy-Efficient Regenerative Braking in Electric Vehicles using Interpretable Recurrent Proximal Policy Optimization

    by Mohammed Zaid Ghawy, Shuyan Chen, Rajasekhar Balasubramanian, Adam A. Q. Mohammed, Yongfeng Ma, Aamir Hussain, Sajan Shaikh

    The increasing prevalence of Electric Vehicles (EVs) on highways is making energy-efficient control decision-making essential. However, optimizing regenerative braking (RB) improves energy recovery, extends driving range, and enhances safety and efficiency. This study proposes an explainable deep reinforcement learning (DRL) framework, Recurrent Proximal Policy Optimization based on Long Short-Term Memory (RPPO-LSTM), to optimize RB in real-world driving conditions. The proposed shared LSTM-based Actor-Critic architecture captures temporal dependencies in high-dimensional driving sequences, enabling efficient policy learning while enhancing regenerative energy recovery. Firstly, a custom OpenAI Gym environment was designed to model realistic EV braking dynamics. Secondly, the model was trained and validated on the publicly available, real-world, downloadable dynamometer dataset (D3). The proposed RPPO-LSTM framework demonstrated superior learning performance, with lower training loss than the PPO strategy, and achieved a 22% improvement in cumulative regenerative power compared with the baseline, while reducing comfort and friction losses and maintaining smooth braking control. RPPO-LSTM achieved the lowest average of total loss (0.0113), value function loss (0.0581), entropy loss (-1.188), Kullback–Leibler (0.00144) and Policy Gradient Loss (-0,005) compared to the PPO strategy. Finally, two complementary interpretability methods were employed: Mutual Information (MI) to select relevant input control signals prior to training, and SHAP to explain the trained model’s decisions post-training and quantify the influence of driving dynamics on energy regeneration. This approach outperforms conventional control strategies by enabling adaptive, real-time energy recovery while maintaining safety and mechanical integrity, showing that DRL can develop more intelligent RB systems for future and autonomous EVs.

  • Open Access

    Hybrid FEM–machine learning framework for back-analysis of spatially varying soil parameters in super-large caisson foundation

    by Dhyaa A.H. Abualghethe, Baogang Mu, Guoliang Dai, Zhongwei Li, Adam A.Q. Mohammed, Amr M.A. Moussa


  • A citation recommendation model employing knowledge graph embedding

    by Zafar Ali, Guilin Qi, Sumaira Hussain, Irfan Ullah, Shah Khalid, Adam A. Q. Mohammed, Inam Ullah, Aalia Malik, Pavlos Kefalas


  • Open Access

    Frequency-specific alterations in low-frequency functional connectivity in children with ADHD

    by Ahmed Ameen Fateh, Hassan Muhammed, Adam A. Q. Mohammed, Abla Smahi, Abdulqawi Alarefi, Luning Zhang, Feng Yue, Hongwu Zeng

    This investigation seeks to elucidate frequency-specific alterations in resting-state functional connectivity (FC) among pediatric subjects with attention-deficit/hyperactivity disorder (ADHD). The study focuses on three distinct sub-bands of slow blood oxygen level-dependent (BOLD) oscillations: Slow3 (0.073–0.198 Hz), Slow4 (0.027–0.073 Hz), and Slow5 (0.010–0.027 Hz). We assess the discriminative capacity of these sub-bands as potential neural biomarkers through the application of machine learning classification techniques. This frequency-resolved methodology addresses a significant lacuna in extant ADHD FC research, which has predominantly treated low-frequency fluctuations as a homogeneous band. By parsing these oscillations into discrete frequency ranges, we aim to provide a more nuanced understanding of the neurophysiological underpinnings of ADHD. This study explores the frequency-specific functional connectivity (FC) differences between (ADHD = 40) and healthy control (HC = 45) children, focusing on the Slow3 (0.073–0.198 Hz), Slow4 (0.027–0.073 Hz), and Slow5 (0.010–0.027 Hz) frequency bands. A two-sample t-test was used to identify significant functional connectivity (FC) differences between the two groups. followed by machine learning techniques to evaluate the classification accuracy of these features. Significant differences in FC were found between ADHD and HC in specific brain regions. In the Slow3 band, ADHD subjects showed increased connectivity in the right precentral gyrus, which is associated with motor-related neural activity. In the Slow4 and Slow5 bands, higher connectivity was observed in the right inferior frontal orbital region in ADHD compared to HC. Given the identical FC patterns in Slow4 and Slow5, these bands were combined into a single feature for further analysis. Machine learning classification using FC features from Slow3 and the combined Slow4/Slow5 bands demonstrated notable accuracy, with 79% accuracy for ADHD and 82% for HC in the Slow3 band, and 85% accuracy for ADHD and 80% for HC in the combined Slow4/Slow5 band. ROC curve analysis further supported these findings with AUC values of 0.7550 and 0.7830, respectively. These findings suggest that frequency-specific FC patterns may serve as potential biomarkers for ADHD, offering insights into the disorder’s neural mechanisms and aiding in more accurate diagnosis. Not applicable.


  • Dynamic functional connectivity variability in the thalamocortical circuit: Insights from Self-Limited Epilepsy with Centrotemporal Spikes (SeLECTs)

    by Ahmed Ameen Fateh, Abla Smahi, Muhammad Hassan, Cristina Cañete-Massé, Adam A.Q. Mohammed, Feng Yue, Zhanqi Hu, Hongwu Zeng


  • Adaptive SimSiam Self-Supervised Networks with Hybrid PIO for Hyperspectral Image Classification

    by Houari Youcef Moudjib, Haibin Duan, Adam A. Q. Mohammed

    Hyperspectral imaging (HSI) provides rich spectral information across hundreds of bands, offering significant potential for fine-grained classification in remote sensing. However, the high dimensionality of HSI data and the scarcity of labeled samples remain persistent challenges. In this paper, we present ASSIAM, a unified framework that combines Hybrid Pigeon-Inspired Optimization (PIO) for spectral band selection with an Adaptive SimSiam self-supervised learning strategy. The Hybrid-PIO mechanism efficiently identifies informative bands by coupling global heuristic search with variance-preserving reduction, while the Adaptive SimSiam model—supported by transfer learning—learns robust spectral–spatial representations from unlabeled data. Extensive experiments on six benchmark HSI datasets, including Salinas, Pavia University, and Indian Pines, demonstrate that ASSIAM achieves consistent improvements over existing methods, with overall classification accuracies reaching up to 99.01%. These results highlight the benefits of integrating bio-inspired optimization with contrastive self-supervised learning to develop scalable and label-efficient solutions for hyperspectral image classification.