Search Results (79)

Distributed learning (DL), in which multiple nodes in an inner-connected network collaboratively induce a predictive model using their local data and some information communicated across neighboring nodes, has received significant research interest in recent years. Yet, it is challenging to achieve excellent performance in scenarios when training data instances have incomplete features and ambiguous labels. In such cases, it is essential to develop an efficient method to jointly perform the tasks of missing feature imputation and credible label recovery. Considering this, in this article, a distributed partial label missing data classification (dPMDC) algorithm is proposed. In the proposed algorithm, an integrated framework is formulated, which takes the ideas of both generative and discriminative learning into account. Firstly, by exploiting the weakly supervised information of ambiguous labels, a distributed probabilistic information-theoretic imputation method is designed to distributively fill in the missing features. Secondly, based on the imputed feature vectors, the classifier modeled by the random feature map of the ${\chi }^2$ kernel function can be learned. Two iterative steps constitute the dPMDC algorithm, which can be used to handle dispersed, distributed data with partially missing features and ambiguous labels. Experiments on several datasets show the superiority of the suggested algorithm from many viewpoints. Full article

Background: This study evaluated a custom algorithm that sought to perform a radiogenomic analysis on lung cancer genetic and imaging data, specifically by using machine learning to see whether a custom clustering/classification method could simultaneously identify features from imaging data that correspond to genetic markers. Methods: CT imaging data and genetic mutation data for 281 subjects with NSCLC were collected from the CPTAC-LUAD and TCGA-LUSC databases on TCIA. The algorithm was run as follows: (1) genetic clusters were initialized using random clusters, binary matrix factorization, or k-means; (2) image classification was run on CT data for these genetic clusters; (3) misclassified subjects were re-classified based on the image classification algorithm; and (4) the algorithm was run until an accuracy of 90% or no improvement after 10 runs. Input genetic mutations were evaluated for potential medical treatments and severity to provide clinical relevance. Results: The image classification algorithm was able to achieve a >90% accuracy after nine algorithm runs and grouped subjects from a starting five clusters to four final clusters, where final image classification accuracy was better than every initial clustered accuracy. These clusters were stable across all three test runs. A total of thirty-eight genes from the top hundred across each subject were identified with specific severity or treatment data; twelve of these genes are listed. Conclusion: This small pilot study presented a potential way to identify genetic patterns from image data and presented a methodology that could group images with no labels or only partial labels for future problems. Full article

The changes in operating conditions of a power transformer can cause a shift in the distribution of partial discharge data, leading to the gradual generation of unlabeled new data, which results in the degradation of the original partial discharge detection model and a decline in its classification performance. To address the aforementioned challenge, a domain-adaptive transformer partial discharge recognition method combining AlexNet-KAN with DANN is proposed. First, the Kolmogorov–Arnold Network (KAN) is introduced to improve the AlexNet model, resulting in the AlexNet-KAN model, which improves the accuracy of transformer partial discharge recognition. Second, the domain adversarial mechanism from domain adaptation theory is applied to the domain of transformer partial discharge recognition, leading to the development of a domain-adaptive transformer partial discharge recognition model that combines AlexNet-KAN with Domain Adversarial Neural Networks (DANNs). Experimental outcomes show that the proposed model effectively adapts transformer partial discharge data from the source domain to the target domain, addressing the issue of distribution shift in transformer partial discharge data with either no labels or very few labels in the new data. Full article

Electricity theft, emerging as one of the severe cyberattacks in smart grids, causes significant economic losses. Due to the powerful expressive ability of deep neural networks (DNN), supervised and unsupervised DNN-based electricity theft detection (ETD) schemes have experienced widespread deployment. However, existing works have the following weak points: Supervised DNN-based schemes require abundant labeled anomalous samples for training, and even worse, cannot detect unseen theft patterns. To avoid the extensively labor-consuming activity of labeling anomalous samples, unsupervised DNNs-based schemes aim to learn the normality of time-series and infer an anomaly score for each data instance, but they fail to capture periodic features effectively. To address these challenges, this paper proposes a novel periodicity-enhanced deep one-class classification framework (PE-DOCC) based on a periodicity-enhanced transformer encoder, named Periodicformer encoder. Specifically, within the encoder, a novel criss-cross periodic attention is proposed to capture both horizontal and vertical periodic features. The Periodicformer encoder is pre-trained by reconstructing partially masked input sequences, and the learned latent representations are then fed into a one-class classification for anomaly detection. Extensive experiments on real-world datasets demonstrate that our proposed PE-DOCC framework outperforms state-of-the-art unsupervised ETD methods. Full article

Recently, the application of deep neural networks to detect anomalies on medical images has been facing the appearance of noisy labels, including overlapping objects and similar classes. Therefore, this study aims to address this challenge by proposing a unique attention module that can assist deep neural networks in focusing on important object features in noisy medical image conditions. This module integrates global context modeling to create long-range dependencies and local interactions to enable channel attention ability by using 1D convolution that not only performs well with noisy labels but also consumes significantly less resources without any dimensionality reduction. The module is then named Global Context and Local Interaction (GCLI). We have further experimented and proposed a partial attention strategy for the proposed GCLI module, aiming to efficiently reduce weighted redundancies. This strategy utilizes a subset of channels for GCLI to produce attention weights instead of considering every single channel. As a result, this strategy can greatly reduce the risk of introducing weighted redundancies caused by modeling global context. For classification, our proposed method is able to assist ResNet34 in achieving up to 82.5% accuracy on the Chaoyang test set, which is the highest figure among the other SOTA attention modules without using any processing filter to reduce the effect of noisy labels. For object detection, the GCLI is able to boost the capability of YOLOv8 up to 52.1% mAP50 on the GRAZPEDWRI-DX test set, demonstrating the highest performance among other attention modules and ranking second in the mAP50 metric on the VinDR-CXR test set. In terms of model complexity, our proposed GCLI module can consume fewer extra parameters up to 225 times and has inference speed faster than 30% compared to the other attention modules. Full article

Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains a leading cause of global mortality. While TB detection can be performed through chest X-ray (CXR) analysis, numerous studies have leveraged AI to automate and enhance the diagnostic process. However, existing approaches often focus on partial or incomplete lesion detection, lacking comprehensive multi-class and multi-label solutions for the full range of TB-related anomalies. To address this, we present a hybrid AI model combining vision transformer (ViT) and convolutional neural network (CNN) architectures for efficient multi-class and multi-label classification of 14 TB-related anomalies. Using 133 CXR images from Dr. Cipto Mangunkusumo National Central General Hospital and 214 images from the NIH datasets, we tackled data imbalance with augmentation, class weighting, and focal loss. The model achieved an accuracy of 0.911, a loss of 0.285, and an AUC of 0.510. Given the complexity of handling not only multi-class but also multi-label data with imbalanced and limited samples, the AUC score reflects the challenging nature of the task rather than any shortcoming of the model itself. By classifying the most distinct TB-related labels in a single AI study, this research highlights the potential of AI to enhance both the accuracy and efficiency of detecting TB-related anomalies, offering valuable advancements in combating this global health burden. Full article

Forest ecosystems play an essential role in ecological balance, supporting biodiversity and climate change mitigation. These ecosystems are crucial not only for ecological stability but also for the local economy. Performing a tree census at a country scale via traditional methods is resource-demanding, error-prone, and requires significant effort by a large number of experts. While emerging technologies such as satellite imagery and AI provide the means for achieving promising results in this task with less effort, considerable effort is still required by experts to annotate hundreds or thousands of images. This study introduces a novel methodology for a tree census classification system which leverages historical and partially labeled data, employs probabilistic data imputation and a weakly supervised training technique, and thus achieves state-of-the-art precision in classifying the dominant tree species of Cyprus. A crucial component of our methodology is a ResNet50 model which takes as input high spatial resolution satellite imagery in the visible band and near-infrared band, as well as topographical features. By applying a multimodal training approach, a classification accuracy of 90% among nine targeted tree species is achieved. Full article

In recent decades, the global use of ashes derived from agro-industrial by-products, such as oil palm kernel shells, which are widely cultivated in Colombia and other tropical regions of the world, has increased. However, the application of these ashes in engineering remains limited due to their heterogeneity and variability. This study utilized scanning electron microscopy (SEM) to assess the influence of calcination temperatures, ranging from 500 °C to 1000 °C, as well as the physical processes of cutting, grinding, and crushing, on the silica content of the studied ashes. Specifically, the sample labeled M18A-c-m-T600°C-t1.5h-tr1h, which was subjected to a calcination temperature of 600 °C and underwent cutting and grinding before calcination, followed by post-calcination crushing, exhibited the highest silica concentration. Complementary techniques such as X-ray fluorescence (XRF) and X-ray diffraction (XRD), were applied to this sample to evaluate its feasibility as an additive or partial replacement for cement in concrete. XRF analysis revealed a composition of 71.24% SiO₂, 9.39% Al₂O₃, and 2.65% Fe₂O₃, thus, meeting the minimum oxide content established by ASTM C 618 for the classification as a pozzolanic material. Furthermore, XRD analysis confirmed that the sample M18A-c-m-T600°C-t1.5h-tr1h is in an amorphous state, which is the only state in which silica can chemically react with calcium hydroxide resulting from the hydration reactions of cement, forming stable cementitious products with strong mechanical properties. Full article

Inthis paper, a distributed semi-supervised partial multi-label learning (dS²PML) algorithm is proposed, which can be used to address the problem of distributed classification of partially multi-labeled data and unlabeled data. In this algorithm, we utilize the multi-kernel function together with the label correlation term to construct the discriminant function. In addition, to obtain a decentralized implementation, we design a reconstructed error on the labeling confidence based on globally common basic data that are selected by a distributed strategy. By exploiting the similarity structure among feature and label spaces under the sparsity constraint, the labeling confidences of partially multi-labeled and unlabeled data are estimated in a decentralized manner. Meanwhile, by using the sparse random feature map to approximate the kernel feature map, the multi-label classifier can be trained under the supervision of the estimated labeling confidence. Experiments on multiple real datasets are conducted to evaluate the learning performance of the proposed approach. According to the experimental results, the average ranks of all the comparison algorithms evaluated on five evaluation metrics are computed. The ranking results show that the average ranks of our algorithm in terms of hamming loss, one error, average precision, ranking loss, and coverage are 3.16, 2.27, 2.15, 2.38, and 2.18, respectively. The average ranks of the dS²PML are second only to the corresponding centralized S²PML (cS²PML) algorithms and higher than other existing comparison algorithms in five evaluation metrics. The average rank differences in terms of Hamming loss, one error, average precision, ranking loss, and coverage between our proposed algorithm and the closest comparison algorithm are 0.28, 1.67, 1.80, 1.15, and 1.62, respectively. Additionally, owing to the distributed storage and decentralized processing of training data, our proposed dS²PML algorithm reduces CPU time by more than 65% and memory consumption by more than 6% compared to the centralized comparison algorithms. The experimental results indicate that our proposed algorithm outperforms the other state-of-the-art algorithms in classification accuracy, CPU time, and memory consumption. Full article

Deep learning approaches have been successfully applied to perform automatic classification of phase-resolved partial discharge (PRPD) diagrams. Under the supervised learning paradigm, however, the performance of classifiers strongly depends on the availability of large and previously labeled data sets. Labeling is an intensive and time-consuming labor, typically involving the manual annotation of a large number of data examples by an expert. In this work, we propose a label propagation algorithm applied to PRPD data sets, aiming to reduce the time necessary to manually label PRPDs. Our basic pipeline is composed of three phases: pre-processing, dimensionality reduction procedures, and clustering. Different configurations of the basic pipeline are tested by using PRPDs obtained from online measurements in hydrogenerators. The performance of each configuration is assessed by using the Silhouette, Caliński–Harabasz, and Davies–Bouldin scores. The clustering of the best three configurations is compared with annotated PRPDs by using the Fowlkes-Mallows index. Results suggest our strategy can substantially decrease the time for manual labeling. Full article

In this paper, a method for augmenting samples of side-scan sonar seafloor sediment images based on CBAM-BCEL1-INGAN is proposed, aiming to address the difficulties in acquiring and labeling datasets, as well as the insufficient diversity and quantity of data samples. Firstly, a Convolutional Block Attention Module (CBAM) is integrated into the residual blocks of the INGAN generator to enhance the learning of specific attributes and improve the quality of the generated images. Secondly, a BCEL1 loss function (combining binary cross-entropy and L1 loss functions) is introduced into the discriminator, enabling it to focus on both global image consistency and finer distinctions for better generation results. Finally, augmented samples are input into an AlexNet classifier to verify their authenticity. Experimental results demonstrate the excellent performance of the method in generating images of coarse sand, gravel, and bedrock, as evidenced by significant improvements in the Frechet Inception Distance (FID) and Inception Score (IS). The introduction of the CBAM and BCEL1 loss function notably enhances the quality and details of the generated images. Moreover, classification experiments using the AlexNet classifier show an increase in the recognition rate from 90.5% using only INGAN-generated images of bedrock to 97.3% using images augmented using our method, marking a 6.8% improvement. Additionally, the classification accuracy of bedrock-type matrices is improved by 5.2% when images enhanced using the method presented in this paper are added to the training set, which is 2.7% higher than that of the simple method amplification. This validates the effectiveness of our method in the task of generating seafloor sediment images, partially alleviating the scarcity of side-scan sonar seafloor sediment image data. Full article

This study examines the performance of various machine learning (ML) models in predicting Interstitial Glucose (IG) levels using data from wrist-worn wearable sensors. The insights from these predictions can aid in understanding metabolic syndromes and disease states. A public dataset comprising information from the Empatica E4 smart watch, the Dexcom Continuous Glucose Monitor (CGM) measuring IG, and a food log was utilized. The raw data were processed into features, which were then used to train different ML models. This study evaluates the performance of decision tree (DT), support vector machine (SVM), Random Forest (RF), Linear Discriminant Analysis (LDA), K-Nearest Neighbors (KNN), Gaussian Naïve Bayes (GNB), lasso cross-validation (LassoCV), Ridge, Elastic Net, and XGBoost models. For classification, IG labels were categorized into high, standard, and low, and the performance of the ML models was assessed using accuracy (40–78%), precision (41–78%), recall (39–77%), F1-score (0.31–0.77), and receiver operating characteristic (ROC) curves. Regression models predicting IG values were evaluated based on R-squared values (−7.84–0.84), mean absolute error (5.54–60.84 mg/dL), root mean square error (9.04–68.07 mg/dL), and visual methods like residual and QQ plots. To assess whether the differences between models were statistically significant, the Friedman test was carried out and was interpreted using the Nemenyi post hoc test. Tree-based models, particularly RF and DT, demonstrated superior accuracy for classification tasks in comparison to other models. For regression, the RF model achieved the lowest RMSE of 9.04 mg/dL with an R-squared value of 0.84, while the GNB model performed the worst, with an RMSE of 68.07 mg/dL. A SHAP analysis identified time from midnight as the most significant predictor. Partial dependence plots revealed complex feature interactions in the RF model, contrasting with the simpler interactions captured by LDA. Full article

Food adulteration which is economically motivated (i.e., food fraud) is an incentive for the development and application of new and fast detection methods/instruments. An example of a fast method that is extremely environmentally friendly is near-infrared spectroscopy (NIRS). Therefore, the goal of this research was to examine the potential of its application in monitoring the adulteration of blended sunflower/olive oils and to compare two types of NIRS instruments, one of which is a portable micro-device, which could be used to assess the purity of olive oil anywhere and would be extremely useful to inspection services. Both NIR devices (benchtop and portable) enable absorbance monitoring in the wavelength range from 900 to 1700 nm. Extra virgin oils (EVOOs) and “ordinary” olive oils (OOs) from large and small producers were investigated, which were diluted with sunflower oil in proportions of 1–15%. However, with the appearance of different salad oils that have a defined share of EVOO stated on the label (usually 10%), the possibilities of the recognition and manipulation in these proportions were tested; therefore, EVOO was also added to sunflower oil in proportions of 1–15%. The composition of fatty acids, color parameters, and total dissolved substances and conductivity for pure and “adulterated” oils were monitored. Standard tools of multivariate analysis were applied, such as (i) analysis of main components for the qualitative classification of oil and (ii) partial regression using the least square method for quantitative prediction of the proportion of impurities and fatty acids. Qualitative models proved successful in classifying (100%) the investigated oils, regardless of whether the added thinner was olive or sunflower oil. Developed quantitative models relating measured parameters with the NIR scans, resulted in values of R² ≥ 0.95 and was reliable (RPD > 8) for fatty acid composition prediction and for predicting the percentage of the added share of impurity oils, while color attributes were less successfully predicted with the portable NIR device (RPD in the range of 2–4.2). Although with the portable device, the prediction potentials remained at a qualitative level (e.g., color parameters), it is important to emphasize that both devices were tested not only with EVOO but also with OO and regardless of whether proportions of 1–15% sunflower oil were added to EVOO and OO or EVOO and OO in the same proportions to sunflower oil. Full article

The Iberian pig is a native breed of the Iberian Peninsula, which holds an international reputation due to the superior quality and the added value of its products. Different rearing practices and feeding regimes are regulated, resulting in different labelling schemes. However, there is no official analytical methodology that is standardised for certification purposes in the sector. Near Infrared Spectroscopy (NIRS) is a technology that provides information about the physicochemical composition of a sample, with several advantages that have enabled its implementation in different fields. Although it has already been successfully used for the analysis of Iberian pig’s final products, samples evaluated with NIRS technology are characterised by a postmortem collection. The goal of this study was to evaluate the potential of NIRS analysis of faeces for in vivo discrimination of the Iberian pig feeding regime, using the spectral information per se for the development of modified partial least squares regressions. Faecal samples were used due to their easy collection, especially in extensive systems where pig handling is difficult. A total of 166 individual samples were collected from 12 farms, where the three different feeding regimes available in the sector were ensured. Although slight differences were detected depending on the chemometric approach, the best models obtained a classification success and a prediction accuracy of over 94% for feeding regime discrimination. The results are considered very satisfactory and suggest NIRS analysis of faeces as a promising approach for the in vivo discrimination of the Iberian pigs’ diet, and its implementation during field inspections, a significative achievement for the sector. Full article

Hematological morphology examinations, essential for diagnosing blood disorders, increasingly utilize deep learning. Blood cell classification, determined by combinations of cell attributes, is complicated by the complex relationships and subtle differences among the attributes, resulting in significant time and cost penalties. This study introduces the Partial Label Learning for Blood Cell Classification (P4BC) strategy, a method that trains neural networks using the blood cell attribute labeling data of weak annotations. Using morphological knowledge, we predefined candidate label sets for the blood cell attributes to blend this knowledge with deep learning. This improves the model’s prediction accuracy and interpretability in classifying attributes. This method effectively combines morphological knowledge with deep learning, an approach we refer to as knowledge alignment. It results in an 8.66% increase in attribute recognition accuracy and a 1.09% improvement in matching predictions to the candidate label sets, compared to the original method. These results confirm our method’s ability to grasp the characteristic information of blood cell attributes, enhancing the model interpretability and achieving knowledge alignment between hematological morphology and deep learning. Our algorithm ensures attribute classification accuracy and shows excellent cell category classification, highlighting its wide application potential and practical value in blood cell category classification. Full article