Search Results (177)

The sixth generation (6G) of wireless networks demands extreme energy efficiency and massive connectivity, positioning large intelligent surfaces (LIS) as a pivotal technology. However, the practical deployment of LIS is constrained by the overwhelming computational complexity and power consumption required to process thousands of antenna elements. To address these challenges, this article proposes a deep learning-based receiver architecture that integrates the spatial efficiency of Partial LIS with advanced non-linear detection. By activating only a subset of antenna panels closest to the user terminal (Partial LIS), the system significantly reduces hardware overhead and Radio Frequency (RF) power consumption. To compensate for the performance loss, the multi-user interference (MUI) generated by the linear combining stage, and the increased MUI inherent in a reduced-aperture environment, a specialized Multilayer Perceptron (MLP) network is implemented. Unlike traditional Zero-Forcing (ZF) or Minimum Mean Squared Error (MMSE) receivers, which require energy-intensive matrix inversions for each frequency component, the proposed neural-network-enabled receiver achieves near-optimal performance using low-complexity combining followed by intelligent learning-based interference suppression. Simulation results demonstrate that the proposed hybrid architecture provides a scalable, “green” solution for 6G uplink scenarios. Notably, the deep learning approach is shown to effectively suppress the performance loss of reduced apertures, achieving a BER comparable to traditional linear benchmarks even with a reduced physical aperture, maintaining good Bit Error Rate (BER) performance while dramatically reducing the computational and hardware footprint. Full article

Accurate performance prediction of variable-speed centrifugal chillers is important for building energy optimization and the development of digital twins in HVAC systems. In practice, obtaining extensive operational data is costly, creating a prevalent “small-sample” dilemma under which conventional data-driven models are prone to overfitting with poor extrapolation capability. While recent Physics-Informed Neural Networks (PINNs) incorporate system-level thermodynamic constraints (e.g., COP definitions), they typically treat the centrifugal compressor as a thermodynamic black box, neglecting its inherent fluid dynamic characteristics; consequently, extrapolated predictions may be physically inconsistent or fall into unsafe operating regions such as compressor surge. To address this gap, this paper proposes an Aero-thermodynamic Physics-Informed Neural Network (Aero-PINN) that introduces three mechanisms into the PINN loss function: (1) dimensionless aerodynamic similarity mapping governed by affinity laws, (2) a surge boundary constraint that prevents non-physical extrapolations, and (3) an aerodynamic–electrical energy coupling validation. Experimental validation on 420 real-world variable-speed test records shows that the Aero-PINN achieves a COP RMSE of 0.04 and a COP MAPE of 0.3%, outperforming standard MLP and polynomial baselines. Moreover, 100% of the extrapolated operating points satisfy all fluid dynamic safety and energy efficiency constraints. This framework provides a reliable, physics-constrained small-sample learning approach, facilitating factory calibration and reduced-test digital modeling for chiller plants. Full article

Background/Objectives: Rapid and accurate classification of ring-enhancing brain lesions (REBLs) into infection or neoplasm is key to clinical triaging for expedited diagnostics in the former to enhance treatment outcomes, especially in the immunocompromised patients. High-resolution three-dimensional (3D) T1 post-contrast (T1+C) MRI provides high-dimensional volumetric data for radiomics analysis. While radiomics is useful in brain neoplasm characterization, its utility in central nervous system infection remains under-explored. In this exploratory study, we aim to determine if a radiomics-machine learning model, based solely on a 3D T1+C MRI dataset, can distinguish infective from neoplastic REBLs. Methods: 92 patients (infection, n = 26; neoplasm, n = 66) with 402 REBLs, who fulfilled criteria for “definite” or “probable” infective or neoplastic REBLs, were identified from scans performed at our hospital over four years and formed the training/validation dataset. All REBLs were manually annotated on T1+C MRI images under radiological supervision. In total, 1197 radiomics features were extracted, feature selection performed using mutual information, and nine machine learning classifiers applied to assess patient-level infection vs. neoplasm classification performance. End-to-end 2D CNN baselines and hybrid radiomics–CNN configurations were additionally evaluated under the same protocol for comparative benchmarking. Model performance was tested on an external holdout dataset of 57 patients (infection, n = 25; neoplasm, n = 32) with 454 REBLs from another hospital. Results: The Multi-layer Perceptron (MLP) model using the Original + LoG + Wavelet feature group demonstrated superior performance. In the cross-validation cohort, it achieved a mean AUC of 0.80 ± 0.02, sensitivity of 0.83 ± 0.09, specificity of 0.77 ± 0.08, and balanced accuracy of 0.80 ± 0.02. On external holdout data, the same configuration showed stable and sustainable performance with an AUC of 0.84, sensitivity of 0.84, specificity of 0.75, and balanced accuracy of 0.80. Conclusions: Our radiomics-machine learning model, based solely on a high-resolution 3D T1+C dataset, shows potential for distinguishing infective REBLs from neoplastic REBLs. Further study, with additional MR sequences and clinical data in a multimodal MRI radiomics-machine learning model, is warranted. Full article

Fault diagnosis in modern power systems increasingly depends on unstructured operation and maintenance (O&M) logs, yet real-world logs are often small in scale and highly imbalanced across fault types, which degrades the generalizability of standard neural models. This paper proposes a graph-augmented diagnostic framework that integrates imbalance-aware knowledge extraction with interpretable reasoning. The framework consists of three stages: (1) domain adaptation of a BERT–BiLSTM–CRF NER model and a BERT–MLP RE model using an imbalance-aware training recipe that combines Low-Rank Adaptation (LoRA), a mixed focal–range loss, and undersampling; (2) construction of a power-system knowledge graph that organizes extracted entities and relations (e.g., fault devices, abnormal phenomena, causes, and handling measures); and (3) a graph-augmented assistant agent that reuses the NER model as a graph-aware retriever within a retrieval-augmented generation (RAG) architecture to support contextualized and interpretable diagnostic reasoning. Experiments on 3921 real-world fault-processing logs show consistent gains: NER reaches 92.0% accuracy and 71.3% Macro-F1 (vs. 80.3% and 63.2%), and RE achieves 88.0% accuracy and 70.1% F1 (vs. 82.1% and 60.4%), while reducing average training time per epoch by about 18%. These results demonstrate an efficient and practical path toward robust log-based fault diagnosis under scarce and imbalanced data. Full article

This study examines information recovery under structured partial observation in multi-informant questionnaire systems. Rather than predicting an external ground truth, we evaluate the recoverability of an operational full-information decision rulewhen only partial informant views are available. In the empirical SNAP-IV calibration study, this reference is intentionally defined as a deterministic function of the combined informant views, so the combined-view result is treated only as an oracle-style ceiling and the substantive analysis concerns how single-view recovery degrades when one informant is withheld. To examine whether a similar qualitative pattern extends beyond this calibration setting, we additionally evaluate a latent-state simulation in which the reference decision is generated from an unobserved latent state and informant views are noisy observations. Across both settings, single-view recoverability declines as inter-rater disagreement increases, whereas combined-view representations remain more stable. In the empirical study, combined-view models achieved near-ceiling recovery performance (e.g., 90.9% for Logistic Regression and 91.3% for MLP), while Teacher-only recovery dropped from approximately 78% to 63% under higher disagreement ($p=0.0005$, Cohen’s $d=1.9$). Additional non-learned single-rater score-threshold baselines exhibited the same qualitative degradation pattern, indicating that the effect is not specific to fitted machine learning models. Importantly, this work is methodological: it does not propose new learning algorithms or clinical prediction models, but instead presents a conceptual–methodological framework, together with model-agnostic recoverability quantities, for quantifying missing-view information loss under incomplete, heterogeneous observations. Full article

This paper investigated the feasibility of snapshot multispectral fluorescence imaging for nondestructive identification of cold stress in pepper plants. Fluorescence spectra were obtained by exciting the plant with a 405 nm ultraviolet LED. The plants were grown under three temperature conditions: 17 °C (control), 10 °C (moderate cold stress), and 5 °C (severe cold stress). Raw fluorescence spectra extracted from the demosaiced snapshot images were used as inputs for a deep-learning pipeline consisting of feature extraction, an encoder–decoder GRU, and a multilayer perceptron (MLP), and the results were compared with conventional machine learning classifiers, including linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and a Gaussian support vector machine (G-SVM). Tukey’s HSD test indicated that the proposed deep-learning model achieved the highest cross-validation accuracy and consistently produced superior classification metrics (accuracy of 85.7%, precision of 85.3%, recall of 85.3%, F1-score of 85.2). The trained model was further applied to hyperspectral cubes to generate classification maps; however, moderate misclassification was observed, consistent with the overall prediction performance. Full article

The burnup, initial enrichment, and cooling time of spent nuclear fuel collectively determine the activities of key gamma-emitting nuclides (e.g., ¹³⁴Cs, ¹³⁷Cs, ¹⁵⁴Eu). In safeguards verification, a non-destructive assay (NDA) using radiation detectors can directly acquire the gamma-ray emission signatures associated with these characteristic nuclides. Previous studies have reported empirical relationships between the activities of nuclides such as ¹³⁴Cs, ¹³⁷Cs, and ¹⁵⁴Eu and the assembly burnup. However, the non-uniform axial power distribution in fuel assemblies leads to variations in axial-segment burnup. Accordingly, this study utilizes a nuclide sample database of a typical pressurized water reactor (PWR) assembly generated by OpenMC 0.15.3 depletion calculations. The calculated results are analyzed, and a sensitivity analysis of the hydrogen-to-uranium atomic ratio (H/U) on the characteristic nuclides is presented, confirming the necessity of incorporating the H/U ratio as an input parameter to improve the cross-condition generalization of the surrogate models. Subsequently, MLP and CNN based on PyTorch 2.9.1 (CUDA 13.0 build: 2.9.1+cu130), and XGBoost 3.0.2 models are implemented to invert axial-segment burnup, initial enrichment, and the number densities of selected actinides under various discrete operating conditions based on characteristic nuclide activities. A comparative analysis of the prediction results from different feature inversion methods is provided. The results indicate that the MLP model performs best with Method A, which incorporates absolute ¹³⁷Cs activity and the ¹⁵⁴Eu/¹³⁷Cs ratio, achieving a relative prediction deviation of only 5.2% for initial enrichment. Under Method C, the XGBoost model attains a relative prediction deviation of only 0.9% for axial-segment burnup (BU_zone). Full article

Human activity recognition (HAR) is essential in many applications, such as smart homes, assisted living, healthcare monitoring, rehabilitation, physiotherapy, and geriatric care. Conventional methods of HAR use wearable sensors, e.g., acceleration sensors and gyroscopes. However, they are limited by issues such as sensitivity to position, user inconvenience, and potential health risks with long-term use. Optical camera systems that are vision-based provide an alternative that is not intrusive; however, they are susceptible to variations in lighting, intrusions, and privacy issues. The paper uses an optical method of recognizing human domestic activities based on pose estimation and deep learning ensemble models. The skeletal keypoint features proposed in the current methodology are extracted from video data using PoseNet to generate a privacy-preserving representation that captures key motion dynamics without being sensitive to changes in appearance. A total of 30 subjects (15 male and 15 female) were sampled across 2734 activity samples, including nine daily domestic activities. There were six deep learning architectures, namely, the Transformer (Transformer), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Multilayer Perceptron (MLP), One-Dimensional Convolutional Neural Network (1D CNN), and a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) architecture. The results on the hold-out test set show that the CNN–LSTM architecture achieves an accuracy of 98.78% within our experimental setting. Leave-One-Subject-Out cross-validation further confirms robust generalization across unseen individuals, with CNN–LSTM achieving a mean accuracy of 97.21% ± 1.84% across 30 subjects. The results demonstrate that vision-based pose estimation with deep learning is a useful, precise, and non-intrusive approach to HAR in smart healthcare and home automation systems. Full article

Connected and Autonomous Vehicles (CAVs) are exposed to increasingly sophisticated cyber threats hidden within high-dimensional, heterogeneous network traffic. A critical bottleneck in existing Intrusion Detection Systems (IDS) is the feature heterogeneity gap: discrete protocol signatures (e.g., flags, services) and continuous traffic statistics (e.g., flow duration, packet rates) reside in disjoint latent spaces. Traditional deep learning approaches typically rely on naive feature concatenation, which fails to capture the intricate, non-linear semantic dependencies between these modalities, leading to suboptimal performance on long-tail, minority attack classes. This paper proposes HCA-IDS, a novel framework centered on Semantics-Aware Cross-Modal Alignment. Unlike heavy-weight models, HCA-IDS adopts a streamlined Multi-Layer Perceptron (MLP) backbone optimized for edge deployment. We introduce a dedicated Multi-Head Cross-Attention mechanism that explicitly utilizes static “Pattern” features to dynamically query and re-weight relevant dynamic “State” behaviors. This architecture forces the model to learn a unified semantic manifold where protocol anomalies are automatically aligned with their corresponding statistical footprints. Empirical assessments on the NSL-KDD and CICIDS2018 datasets, validated through rigorous 5-Fold Cross-Validation, substantiate the robustness of this approach. The model achieves a Macro-F1 score of over 94% on 7 consolidated attack categories, exhibiting exceptional sensitivity to minority attacks (e.g., Web Attacks and Infiltration). Crucially, HCA-IDS is ultra-lightweight, with a model size of approximately 1.00 MB and an inference latency of 0.0037 ms per sample. These results confirm that explicit semantic alignment combined with a lightweight architecture is key to robust, real-time intrusion detection in resource-constrained CAVs. Full article

In complex natural scenarios, facial expression recognition often encounters partial occlusions caused by glasses, hand gestures, and hairstyles, making it difficult for models to extract effective features and thereby reducing recognition accuracy. Existing methods often employ attention mechanisms to enhance expression-related features, but they fail to adequately address the issue where high-frequency responses in occluded regions can disperse attention weights (e.g., incorrectly focus on occluded areas), making it challenging to effectively utilize local cues around the occlusions and limiting performance improvement. To address this, this paper proposes a network based on an adaptive attention mechanism (Adaptive Attention Vision Transformer, AA-ViT). First, an Adaptive Attention module (ADA) is designed to dynamically adjust attention scores in occluded regions, enhancing the effective information in features. Next, a Dual-Branch Multi-Layer Perceptron (DB-MLP) replaces the single linear layer to improve feature representation and model classification capability. Additionally, a Random Erasure (RE) strategy is introduced to enhance model robustness. Finally, to address the issue of model training instability caused by class imbalance in the training dataset, a hybrid loss function combining Focal Loss and Cross-Entropy Loss is adopted to ensure training stability. Experimental results show that AA-ViT achieves expression recognition accuracies of 90.66% and 90.01% on the RAF-DB and FERPlus datasets, respectively, representing improvements of 4.58 and 18.9 percentage points over the baseline ViT model, with only a 24.3% increase in parameter count. Compared to existing methods, the proposed approach demonstrates superior performance in occluded facial expression recognition tasks. Full article

Advances in artificial intelligence (AI), soft robotics, and miniaturized imaging technologies have accelerated the development of endorobotic platforms that aim to enhance detection accuracy and improve patient experience. In this narrative review, we synthesize evidence on AI-assisted detection and characterization systems (CADe/CADx), robotic locomotion mechanisms, adhesion strategies, imaging modalities, and material and power constraints relating to next-generation CRC screening technologies. Reported performance metrics are interpreted within their original methodological contexts, acknowledging the heterogeneity of datasets, limited representation of diverse populations, underreporting of negative findings, and scarcity of large, real-world comparative trials. We introduce a conceptual translational framework that links engineering design principles with validation needs across in silico, in vitro, preclinical, and clinical stages, and we outline safety considerations, workflow integration challenges, and sterility requirements that influence real-world deployability. Regulatory alignment is discussed using the U.S. FDA Total Product Life Cycle (TPLC) and Good Machine Learning Practice (GMLP) frameworks to highlight expectations for data quality, model robustness, device–software interoperability, and post-market monitoring. Collectively, the evidence demonstrates promising technological innovation but also highlights substantial gaps that must be addressed before AI-enabled endorobotic systems can be safely and effectively integrated into routine CRC screening. Continued interdisciplinary work, supported by rigorous validation and transparent reporting, will be essential to advance these technologies toward meaningful clinical impact. Full article

Chloride diffusivity of concrete is essentially determined by its microstructural parameters. Establishing a reliable and accurate prediction model for chloride diffusion has become a research hotspot. In this study, a database containing 144 sets of macro–micro property parameters of concrete is established to train a Multilayer Perceptron (MLP) model. Taking the original collected data as a benchmark, data are randomly missing to simulate data incompleteness, and the models are trained using data filled by the Lagrange, K-Nearest Neighbor (KNN), and Miceforest methods. Moreover, the original data is expanded by the virtual sample generation (VSG) algorithm, based on a Gaussian mixture model (GMM) that fits the joint probability distribution of the original data to generate virtual samples preserving statistical (mean, standard deviation) and physical (e.g., porosity range, pore size ratio) consistency, thus mitigating the randomness caused by small sample sizes. Results indicate that the MLP model demonstrates excellent predictive performance: among schemes handling missing data, the model preprocessed by normalization with KNN imputation yields the best results with testing R² of 0.78; the baseline model (without missing value filling, normalized) achieves testing R² of 0.83, MAE of 0.572, and MSE of 0.424. VSG-expanded data significantly enhances the MLP model’s prediction accuracy. When expanding to 3000 groups, the testing R² reaches 0.85, a 2.4% increase compared to 1000 groups, with further improvements as the dataset expands, confirming the feasibility of the VSG algorithm for small-sample scenarios. Full article

Numerous valuable information is available on the Internet, and many individuals rely on mass media as their primary source of information. Various views, comments, expressions, and opinions on social networks have been a tremendous source of information. Harvesting free, resourceful information through social media makes text mining a powerful tool for analyzing public opinions on various issues across diverse social networks. Various research projects have implemented text sentiment analysis through machine and deep learning approaches. Social media text often expresses sentiment through complex syntax and negation (e.g., implicit and double negation and nested clauses), which many classifiers mishandle. We propose hybrid negation, a clause-aware approach that combines (i) explicit/implicit/double-negation rules, (ii) dependency-based scope detection, (iii) a TextBlob back-off for phrase polarity, and (iv) an MLP-learned clause-weighting module that aggregates clause-level scores. Across 156,539 tweets (three-class sentiment), we evaluate six negation strategies and 228 model configurations with and without SMOTE (applied strictly within training folds). Hybrid Negation achieves 98.582% accuracy, 98.196% precision, 98.189% recall, and 98.193% F1 with BERT, outperforming rule-only and antonym/synonym baselines. Ablations show each component contributes to the model’s performance, with dependency scope and double negations offering the largest gains. Per-class results, confidence intervals, and paired tests with multiple-comparison control confirm statistically significant improvements. We release code and preprocessing scripts to support reproducibility. Full article

Background/Objectives: Rapid and accurate diagnosis of acute appendicitis is crucial for patient health and management, and the diagnostic process can be prolonged due to varying clinical symptoms and limitations of diagnostic tools. This study aims to shorten the timeframe for these vital processes and increase accuracy by developing a quantum-inspired hybrid model to identify appendicitis types. Methods: The developed model initially selects the two most performing architectures using four convolutional neural networks (CNNs) and two Transformers (ViTs). Feature extraction is then performed from these architectures. Phase-based trigonometric embedding, low-order interactions, and norm-preserving principles are used to generate a Quantum Feature Map (QFM) from these extracted features. The generated feature map is then passed to the Multiple Head Attention (MHA) layer after undergoing Hadamard fusion. At the end of this stage, classification is performed using a multilayer perceptron (MLP) with a ReLU activation function, which allows for the identification of acute appendicitis types. The developed quantum-inspired hybrid model is also compared with six different CNN and ViT architectures recognized in the literature. Results: The proposed quantum-inspired hybrid model outperformed the other models used in the study for acute appendicitis detection. The accuracy achieved in the proposed model was 97.96%. Conclusions: While the performance metrics obtained from the quantum-inspired model will form the basis of deep learning architectures for quantum technologies in the future, it is thought that if 6G technology is used in medical remote interventions, it will form the basis for real-time medical interventions by taking advantage of quantum speed. Full article

Earthquakes are the result of the dynamic processes occurring beneath the Earth’s crust; specifically, the movement and interaction of tectonic/lithospheric plates. When one plate shifts relative to another, stress accumulates and is eventually released as seismic energy. This process is continuous and unstoppable. This phenomenon is well recognized in the Mediterranean region, where significant seismic activity arises from the northward convergence (4–10 mm per year) of the African plate relative to the Eurasian plate along a complex plate boundary. Consequently, our research will focus on the Mediterranean region, specifically examining seismic activity from 1990 to 2015 within the latitude range of 33–44° and longitude range of 17–44°. These geographical coordinates encompass 28 seismic zones, with the most active areas being Turkey and Greece. In this paper, we applied Grammatical Evolution for artificial feature construction in earthquake prediction, evaluated against machine learning approaches including MLP(GEN), MLP(PSO), SVM, and NNC. Experiments showed that feature construction (FC) achieved the best performance, with a mean error of 9.05% and overall accuracy of 91%, outperforming SVM. Further analysis revealed that a single constructed feature $\left(N_f=1\right)$ yielded the lowest average error (8.21%), while varying the number of generations indicated that $N_g=200$ provided an effective balance between computational cost and predictive accuracy. These findings confirm the efficiency of FC in enhancing earthquake prediction models through artificial feature construction. Our results, as will be discussed in greater detail within the research, yield an average error of approximately 9%, corresponding to an overall accuracy of 91%. Full article