Search Results (94)

Conventional sensing expends energy at three stages: powering dedicated sensors, transmitting measurements, and executing computationally intensive inference. Wireless sensing re-purposes WiFi channel state information (CSI) inherent in every packet, eliminating extra sensors and uplink traffic, though reliance on deep neural networks (DNNs) often trained and run on graphics processing units (GPUs) can negate these gains. This review highlights two core energy efficiency levers in CSI-based wireless sensing. First ambient CSI harvesting cuts power use by an order of magnitude compared to radar and active Internet of Things (IoT) sensors. Second, integrated sensing and communication (ISAC) embeds sensing functionality into existing WiFi links, thereby reducing device count, battery waste, and carbon impact. We review conventional handcrafted and accuracy-first methods to set the stage for surveying green learning strategies and lightweight learning techniques, including compact hybrid neural architectures, pruning, knowledge distillation, quantisation, and semi-supervised training that preserve accuracy while reducing model size and memory footprint. We also discuss hardware co-design from low-power microcontrollers to edge application-specific integrated circuits (ASICs) and WiFi firmware extensions that align computation with platform constraints. Finally, we identify open challenges in domain-robust compression, multi-antenna calibration, energy-proportionate model scaling, and standardised joules per inference metrics. Our aim is a practical battery-friendly wireless sensing stack ready for smart home and 6G era deployments. Full article

Network intrusion behaviors exhibit high concealment and diversity, making intrusion detection methods based on single-behavior modeling unable to accurately characterize such activities. To overcome this limitation, we propose SGMT-IDS, a dual-branch semi-supervised intrusion detection model based on Graph Neural Networks (GNNs) and Transformers. By constructing two views of network attacks, namely structural and behavioral semantics, the model performs collaborative analysis of intrusion behaviors from both perspectives. The model adopts a dual-branch architecture. The SGT branch captures the structural embeddings of network intrusion behaviors, and the GML-Transformer branch extracts the semantic information of intrusion behaviors. In addition, we introduce a two-stage training strategy that optimizes the model through pseudo-labeling and contrastive learning, enabling accurate intrusion detection with only a small amount of labeled data. We conduct experiments on the NF-Bot-IoT-V2, NF-ToN-IoT-V2, and NF-CSE-CIC-IDS2018-V2 datasets. The experimental results demonstrate that SGMT-IDS achieves superior performance across multiple evaluation metrics. Full article

Clustering plays a central role in uncovering latent structure within both temporal and visual data. It enables critical insights in various domains including healthcare, finance, surveillance, autonomous systems, and many more. With the growing volume and complexity of time-series and image-based datasets, there is an increasing demand for robust, flexible, and scalable clustering algorithms. Although these modalities differ—time-series being inherently sequential and vision data being spatial—they exhibit common challenges such as high dimensionality, noise, variability in alignment and scale, and the need for interpretable groupings. This survey presents a comprehensive review of recent advancements in clustering methods that are adaptable to both time-series and vision data. We explore a wide spectrum of approaches, including distance-based techniques (e.g., DTW, EMD), feature-based methods, model-based strategies (e.g., GMMs, HMMs), and deep learning frameworks such as autoencoders, self-supervised learning, and graph neural networks. We also survey hybrid and ensemble models, as well as semi-supervised and active clustering methods that leverage minimal supervision for improved performance. By highlighting both the shared principles and the modality-specific adaptations of clustering strategies, this work outlines current capabilities and open challenges, and suggests future directions toward unified, multimodal clustering systems. Full article

The Qinghai–Tibetan plateau is undergoing severe grassland degradation, commonly known as black-soil areas, caused by overgrazing, climate change, and rodent activity. Accurate black-soil area detection is critical for guiding ecological restoration. However, obtaining large-scale annotated datasets is costly due to the ambiguous visual characteristics and high ecological variability of black-soil areas, often necessitating expert validation and repeated refinement. To address this challenge, we propose SBLS (Semi-supervised Black-Soil area detection), a semi-supervised approach that leverages limited labeled data alongside abundant unlabeled imagery. SBLS adopts a cross-branch pseudo supervision strategy, where pseudolabels generated from weakly augmented views in one branch supervise four strongly augmented views in the other branch. To further capitalize on the unlabeled data, we implement a dual-level contrastive learning approach that operates across both low-level and high-level feature spaces of strongly augmented pairs. Experiments demonstrate that SBLS significantly outperforms existing state-of-the-art methods, establishing a new benchmark for black-soil area detection in semi-supervised settings. Full article

This paper addresses the fundamental problem of seed selection in semi-supervised clustering, where the quality of initial seeds has a significant impact on clustering performance and stability. Existing methods often rely on randomly or heuristically selected seeds, which can propagate errors and increase dependence on expert labeling. To overcome these limitations, we propose FA-Seed, a flexible and adaptive model that integrates active querying with self-guided adaptation within the framework of fuzzy hyperboxes. FA-Seed partitions the data into hyperboxes, evaluates seed reliability through measures of membership and association density, and propagates labels with an emphasis on label purity. The model demonstrates strong adaptability to complex and ambiguous data distributions in which cluster boundaries are vague or overlapping. The main contributions of FA-Seed include: (1) automatic estimation and selection of candidate seeds that provide auxiliary supervision, (2) dynamic cluster expansion without retraining, (3) automatic detection and identification of structurally complex regions based on cluster characteristics, and (4) the ability to capture intrinsic cluster structures even when clusters vary in density and shape. Empirical evaluations on benchmark datasets, specifically the UCI and Computer Science collections, show that our approach consistently outperforms several state-of-the-art semi-supervised clustering methods. Full article

This study proposes a new concept that explains a source of technological change: The invasive behaviour of general purpose technologies that breaks into scientific and technological ecosystems with accelerated diffusion of new products and processes that destroy the usage value of all units previously used. This study highlights the dynamics of the invasive destruction of new path-breaking technologies in driving innovative activity. Invasive technologies conquer the scientific, technological, and business spaces of alternative technologies by introducing manifold radical innovations that support technological, economic, and social change. The proposed theoretical framework is verified empirically in new technologies of neural network architectures, comparing transformer technology (a deep learning architecture having unsupervised and semi-supervised algorithms that create new contents and mimic human ability, supporting Generative Artificial Intelligence) to Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNNs). Statistical evidence here, based on patent analyses, reveals that the exponential growth rate of transformer technology over a period of five years (2020–2024) is 45.91% more than double compared to the alternative technologies of LSTM (21.17%) and RNN (18.15%). Moreover, the proposed invasive rate in technological space shows that is very high for transformer technology at the level of 2.2%, whereas for LSTM it is 1.39% and for RNN it is 1.22% over 2020–2024, respectively. Invasive behaviour of drastic technologies is a new approach that can explain one of the major causes of global technological change and this scientific examination here significantly contributes to our understanding of the current dynamics in technological evolution of the Artificial Intelligence technology having high industrial impacts on the progress of human society. Full article

Artificial intelligence (AI) is becoming an integral part of pathological assessment and diagnostic procedures in modern pathology. As most prostate cancers (PCa) arise from glandular epithelial tissue, an AI-based methodology has been developed to recognize glandular epithelial nuclei in prostate biopsy tissue. An integrated machine-learning network, named GlandNet, was developed to correctly recognize the epithelial cells within prostate glands using cell-centric patches selected from the core biopsy specimens. Feulgen-Thionin (a DNA stoichiometric label) was used to stain biopsy sections (4–7 µm in thickness) from 82 active surveillance patients diagnosed with PCa. Images of these sections were human-annotated, and the resultant dataset consisted of 1,264,772 segmented, cell-centric nuclei patches, of which 449,879 were centered on epithelial gland nuclei from 110 needle biopsies (training set: n = 66; validation set: n = 22; and test set: n = 22). The training of GlandNet used semi-supervised machine-learning knowledge of the training and validation cohorts and integrated both human and AI predictions to enhance its performance on the test cohort. The performance was evaluated against a consensus deliberation from three observers. The GlandNet demonstrated an average accuracy, sensitivity, specificity, and F1-score of 94.1%, 95.7%, 87.8%, and 95.2%, respectively, when tested on the 20,735 glandular cells found in the three needle biopsies with the visually best consensus predictions. Conversely, the average accuracy, sensitivity, specificity, and F1-score were 90.9%, 86.4%, 94.0%, and 89.7% when assessed on 57,217 cells found in the three needle biopsies with the visually worst consensus predictions. GlandNet is a first-generation AI with an excellent ability to differentiate between epithelial and stromal nuclei in core biopsies from patients with early prostate cancer. Full article

Satellite imagery provides a rich source of information that serves as a comprehensive and synoptic tool for the continuous monitoring of active volcanoes, including those in remote and inaccessible areas. The huge influx of such data requires the development of automated systems for efficient processing and interpretation. Early warning systems, designed to process satellite imagery to identify signs of impending eruptions and monitor eruptive activity in near real-time, are essential for hazard assessment and risk mitigation. Here, we propose a machine learning approach for the automatic classification of pixels in SEVIRI images to detect and characterize the eruptive activity of a volcano. In particular, we exploit a semi-supervised GAN (SGAN) model that retrieves the presence of thermal anomalies, volcanic ash plumes, and meteorological clouds in each SEVIRI pixel, allowing time series plots to be obtained showing the evolution of volcanic activity. The SGAN model was trained and tested using the huge amount of data available on Mount Etna (Italy). Then, it was applied to other volcanoes, specifically, Stromboli (Italy), Tajogaite (Spain), and Nyiragongo (Democratic Republic of the Congo), to assess the model’s ability to generalize. The validation of the model was performed through a visual comparison between the classification results and the corresponding SEVIRI images. Moreover, we evaluate the model performance by calculating three different metrics, namely the precision (correctness of positive predictions), the recall (ability to find all the positive instances), and the F1-score (general model’s accuracy), finding an average accuracy of 0.9. Our approach can be extended to other geostationary satellite data and applied worldwide to characterize volcanic activity, allowing the monitoring of even remote volcanoes that are difficult to reach from the ground. Full article

Electrophilic compounds are bioactive components commonly found in foods that are capable of covalently modifying nucleophilic sites on biologically functional macromolecules. These compounds may elicit positive bioactivity or negative biotoxicity, posing significant challenges in terms of time and resource expenditure in the de novo characterization of their biological activity. In this study, we developed a database of 332 food-derived electrophilic compounds and used a semi-supervised k-nearest neighbors (KNN) machine learning model to predict their bioactivity. Molecular docking analysis identified the three chalcone compounds with the highest potential positive activity—4-hydroxyderricin (4HD), isoliquiritigenin (ISO), and butein. Furthermore, in cell experiments, treatment with 4HD, ISO, and butein significantly reduced reactive oxygen species (ROS) levels. An RT-qPCR analysis demonstrated that these chalcones significantly upregulated the mRNA expression of Nrf2 and its downstream antioxidant genes, including Nqo1, HO-1, Gsr, Gclc, and Gclm. ISO’s cytoprotective and antioxidant effects were abolished following these findings, which highlight that 4HD, ISO, and butein are effective Nrf2 activators and suggest that comprehensive virtual technology is a promising strategy for identifying functional bioactive compounds. Full article

To efficiently utilize limited resources, this paper proposes a semi-supervised object detection (SSOD) approach based on novel adaptive weighted active learning (AWAL) and orthogonal data augmentation (ODA). An uncertainty sampling framework is applied by adaptively weighting multiple evaluations to annotate the most informative samples for active learning. To further exploit the discriminant potential of unlabeled data, an adaptive weighted loss is introduced to fully mine the unlabeled data, and the normalized uncertainty score is adopted as the loss weight to explore low-score samples for training iterations. Moreover, an ODA operation is performed as pseudo-supervised learning on augmented instances to further capture the modality diversity of complex data distributions. Extensive evaluation and analysis are conducted on the MS-COCO dataset, achieving a mean average precision (mAP) of 35.10 with only 10% of the annotated data. Compared with the existing active learning baselines, the AWAL strategy improves the performance by 1.3% without the ODA. When ODA is incorporated, an additional performance gain of 1.2% is observed. Furthermore, training on the fully annotated MS-COCO with additional unlabeled data, the performance achieved at 43.30 mAP, demonstrating the superiority of the proposed approach. Full article

In open-pit mining, autonomous trucks are essential for enhancing both safety and productivity. Object detection technology is critical to their smooth and secure operation, but training these models requires large amounts of high-quality annotated data representing various conditions. It is expensive and time-consuming to collect these data during open-pit mining due to the harsh environmental conditions. Simulation engines have emerged as an effective alternative, generating diverse labeled data to augment real-world datasets. However, discrepancies between simulated and real-world environments, often referred to as the Sim2Real domain shift, reduce model performance. This study addresses these challenges by presenting a novel semi-supervised domain adaptation for object detection (SSDA-OD) framework named Adamix, which is designed to reduce domain shift, enhance object detection, and minimize labeling costs. Adamix builds on a mean teacher architecture and introduces two key modules: progressive intermediate domain construction (PIDC) and warm-start adaptive pseudo-label (WSAPL). PIDC builds intermediate domains using a mixup strategy to reduce source domain bias and prevent overfitting, while WSAPL provides adaptive thresholds for pseudo-labeling, mitigating false and missed detections during training. When evaluated in a Sim2Real scenario, Adamix shows superior domain adaptation performance, achieving a higher mean average precision (mAP) compared with state-of-the-art methods, with 50% less labeled data required, achieved through active learning. The results demonstrate that Adamix significantly reduces dependence on costly real-world data collection, offering a more efficient solution for object detection in challenging open-pit mining environments. Full article

In the health monitoring of electromechanical transmission systems, the collected state data typically consist of only a minimal amount of labeled data, with a vast majority remaining unlabeled. Consequently, deep learning-based diagnostic models encounter the challenge of scarcity in labeled data and abundance in unlabeled data. Traditional semi-supervised deep learning methods based on pseudo-label self-training, while alleviating the issue of labeled data scarcity to some extent, neglect the reliability of pseudo-label information, the accuracy of feature extraction from unlabeled data, and the imbalance in sample selection. To address these issues, this paper proposes a novel semi-supervised fault diagnosis method under imbalanced unlabeled sample class information screening. Firstly, an information screening mechanism for unlabeled data based on active learning is established. This mechanism discriminates based on the variability of intrinsic feature information in fault samples, accurately screening out unlabeled samples located near decision boundaries that are difficult to separate clearly. Then, combining the maximum membership degree of these unlabeled data in the classification space of the supervised model and interacting with the active learning expert system, label information is assigned to the screened unlabeled data. Secondly, a cost-sensitive function driven by data imbalance is constructed to address the class imbalance problem in unlabeled sample screening, adaptively adjusting the weights of different class samples during model training to guide the training of the supervised model. Ultimately, through dynamic optimization of the supervised model and the feature extraction capability of unlabeled samples, the recognition ability of the diagnostic model for unlabeled samples is significantly enhanced. Validation through two datasets, encompassing a total of 12 experimental scenarios, demonstrates that in scenarios with only a small amount of labeled data, the proposed method achieves a diagnostic accuracy increment exceeding 10% compared to existing typical methods, fully validating the effectiveness and superiority of the proposed method in practical applications. Full article

Power grid operation occurs in complex, dynamic environments where the timely identification of operator violations is essential for safety. Traditional methods often rely on manual supervision and rule-based detection, leading to inefficiencies. Existing deep learning approaches, while powerful, require fully labeled data and long training times, thereby increasing costs. To address these challenges, we propose an active hard sample learning method specifically for the violation action recognition of operators in power grid operation. We design a hard instance sampling module with multi-strategy fusion based on active learning to improve training efficiency. This module identifies hard samples based on the consistency of models or samples, where we develop uncertainty evaluation and the instance discrimination strategy to assess the contributions of samples effectively. We utilize ResNet50 and ViT architectures with Faster-RCNN for detection and recognition, developed using PyTorch 2.0. The dataset comprises 2000 samples, and 30% and 60% labeled data are employed. Experimental results show significant improvements in model performance and training efficiency, demonstrating the method’s effectiveness in complex power grid environments. Our approach enhances safety monitoring and advances active learning and hard sample techniques in practical applications. Full article

The safety and efficiency of assembly lines are critical to manufacturing, but human supervisors cannot oversee all activities simultaneously. This study addresses this challenge by performing a comparative study to construct an initial real-time, semi-supervised temporal action recognition setup for monitoring worker actions on assembly lines. Various feature extractors and localization models were benchmarked using a new assembly dataset, with the I3D model achieving an average mAP@IoU=0.1:0.7 of 85% without optical flow or fine-tuning. The comparative study was extended to self-supervised learning via a modified SPOT model, which achieved a mAP@IoU=0.1:0.7 of 65% with just 10% of the data labeled using extractor architectures from the fully-supervised portion. Milestones include high scores for both fully and semi-supervised learning on this dataset and improved SPOT performance on ANet1.3. This study identified the particularities of the problem, which were leveraged and referenced to explain the results observed in semi-supervised scenarios. The findings highlight the potential for developing a scalable solution in the future, providing labour efficiency and safety compliance for manufacturers. Full article

Fatigue plays a critical role in sports science, significantly affecting recovery, training effectiveness, and overall athletic performance. Understanding and predicting fatigue is essential to optimize training, prevent overtraining, and minimize the risk of injuries. The aim of this study is to leverage Human Activity Recognition (HAR) through deep learning methods for dimensionality reduction. The use of Adversarial AutoEncoders (AAEs) is explored to assess and visualize fatigue in a two-dimensional latent space, focusing on both semi-supervised and conditional approaches. By transforming complex time-series data into this latent space, the objective is to evaluate motor changes associated with fatigue within the participants’ motor control by analyzing shifts in the distribution of data points and providing a visual representation of these effects. It is hypothesized that increased fatigue will cause significant changes in point distribution, which will be analyzed using clustering techniques to identify fatigue-related patterns. The data were collected using a Wii Balance Board and three Inertial Measurement Units, which were placed on the hip and both forearms (distal part, close to the wrist) to capture dynamic and kinematic information. The participants followed a fatigue-inducing protocol that involved repeating sets of 10 repetitions of four different exercises (Squat, Right Lunge, Left Lunge, and Plank Jump) until exhaustion. Our findings indicate that the AAE models are effective in reducing data dimensionality, allowing for the visualization of fatigue’s impact within a 2D latent space. The latent space representation provides insights into motor control variations, revealing patterns that can be used to monitor fatigue levels and optimize training or rehabilitation programs. Full article