Class Labels Research Articles

Early Detection of Verticillium Wilt in Cotton by Using Hyperspectral Imaging Combined with Recurrence Plots

Cotton is susceptible to Verticillium wilt (VW) during its growth. Early and accurate detection of VW can facilitate targeted pesticide treatment and reduce the potential spread of the disease. However, accurately detecting VW in cotton before symptoms appear (the asymptomatic period) after infection by Verticillium dahliae remains challenging. This study proposes an early detection method for cotton wilt disease using hyperspectral imaging and recurrence plots (RP) combined with machine learning techniques. First, spectral curves were collected and analyzed under three conditions of cotton plants: healthy, asymptomatic, and symptomatic. Then, the one-dimensional spectral curve was transformed into two-dimensional recurrence plots to enhance the detail differences in the original spectral curve of cotton plants in various states. Hyperspectral recurrence plots contain rich texture information; fifteen texture features were extracted from the spectral recurrence plots using the Gray-Level Gradient Co-occurrence Matrix (GLGCM). Eleven of these texture features showed a strong correlation with the class labels of the cotton plants. In order to reduce redundant information between features, principal component analysis (PCA) was used to extract the first five principal components, which explained 99.02% of the information from the 11 features. The final principal component dataset was then input into KNN, SVM, ELM, and XGBoost classifiers to assess the accuracy of early detection of VW in cotton. The results showed that the XGBoost model, based on the first five principal components obtained from the texture features, achieved accuracy, precision, recall, and F1-score of 96.3%, 95.6%, 96%, and 95.8%, demonstrating a high classification capability. The results of this study confirm the feasibility of converting spectral curves into recurrence plots and extracting image texture features for the accurate identification of VW in cotton during the asymptomatic period. This method also provides a new strategy for early disease detection of cotton and other plants in the future.

Tomato ripeness and stem recognition based on improved YOLOX

To address the challenges of unbalanced class labels with varying maturity levels of tomato fruits and low recognition accuracy for both fruits and stems in intelligent harvesting, we propose the YOLOX-SE-GIoU model for identifying tomato fruit maturity and stems. The SE focus module was incorporated into YOLOX to improve the identification accuracy, addressing the imbalance in the number of tomato fruits and stems. Additionally, we optimized the loss function to GIoU loss to minimize discrepancies across different scales of fruits and stems. The mean average precision (mAP) of the improved YOLOX-SE-GIoU model reaches 92.17%. Compared to YOLOv4, YOLOv5, YOLOv7, and YOLOX models, the improved model shows an improvement of 1.17–22.21%. The average precision (AP) for unbalanced semi-ripe tomatoes increased by 1.68–26.66%, while the AP for stems increased by 3.78–45.03%. Experimental results demonstrate that the YOLOX-SE-GIoU model exhibits superior overall recognition performance for unbalanced and scale-variant samples compared to the original model and other models in the same series. It effectively reduces false and missed detections during tomato harvesting, improving the identification accuracy of tomato fruits and stems. The findings of this work provide a technical foundation for developing advanced fruit harvesting techniques.

Label credibility correction based on cell morphological differences for cervical cells classification

Cervical cancer is one of the deadliest cancers that pose a significant threat to women’s health. Early detection and treatment are commonly used methods to prevent cervical cancer. The use of pathological image analysis techniques for the automatic interpretation of cervical cells in pathological slides is a prominent area of research in the field of digital medicine. According to The Bethesda System, cervical cytology necessitates further classification of precancerous lesions based on positive interpretations. However, clinical definitions among different categories of lesion are complex and often characterized by fuzzy boundaries. In addition, pathologists can deduce different criteria for judgment based on The Bethesda System, leading to potential confusion during data labeling. Noisy labels due to this reason are a great challenge for supervised learning. To address the problem caused by noisy labels, we propose a method based on label credibility correction for cervical cell images classification network. Firstly, a contrastive learning network is used to extract discriminative features from cell images to obtain more similar intra-class sample features. Subsequently, these features are fed into an unsupervised method for clustering, resulting in unsupervised class labels. Then unsupervised labels are corresponded to the true labels to separate confusable and typical samples. Through a similarity comparison between the cluster samples and the statistical feature centers of each class, the label credibility analysis is carried out to group labels. Finally, a cervical cell images multi-class network is trained using synergistic grouping method. In order to enhance the stability of the classification model, momentum is incorporated into the synergistic grouping loss. Experimental validation is conducted on a dataset comprising approximately 60,000 cells from multiple hospitals, showcasing the effectiveness of our proposed approach. The method achieves 2-class task accuracy of 0.9241 and 5-class task accuracy of 0.8598. Our proposed method achieves better performance than existing classification networks on cervical cancer.

Dynamic Color Assignment for Hierarchical Data.

Assigning discriminable and harmonic colors to samples according to their class labels and spatial distribution can generate attractive visualizations and facilitate data exploration. However, as the number of classes increases, it is challenging to generate a high-quality color assignment result that accommodates all classes simultaneously. A practical solution is to organize classes into a hierarchy and then dynamically assign colors during exploration. However, existing color assignment methods fall short in generating high-quality color assignment results and dynamically aligning them with hierarchical structures. To address this issue, we develop a dynamic color assignment method for hierarchical data, which is formulated as a multi-objective optimization problem. This method simultaneously considers color discriminability, color harmony, and spatial distribution at each hierarchical level. By using the colors of parent classes to guide the color assignment of their child classes, our method further promotes both consistency and clarity across hierarchical levels. We demonstrate the effectiveness of our method in generating dynamic color assignment results with quantitative experiments and a user study.

Fast online feature selection in streaming data

The challenge of getting big amounts of high-quality labeled data is compounded by the fact that data labeling is often subjective and requires significant human effort. In many cases, the quality of the labeled data depends entirely on the expertise and experience of human annotators, making it challenging to ensure labeling accuracy in large and dynamic datasets. Moreover, there may be a significant delay between the arrival of a new instance and its manual labeling. This paper explores the use of fully unsupervised feature selection algorithms in non-stationary data streams, where the importance of features may change over time. We introduce a novel feature selection algorithm called Online Fast FEa-ture SELection-OFFESEL, which calculates the feature importance scores in each incoming window based on their mean normalized values and without using any class labels. We evaluate OFFESEL on 17 benchmark data streams, both stationary and non-stationary, using popular online classifiers like PerceptronMask, VFDT, Online Boosting, and Linear SVM. We compare OFFESEL to several other feature selection algorithms, including state-of-the-art supervised ones like FIRES and ABFS, as well as popular unsupervised ones like MCFS, LS, and Max Variance, which we adapted to data streams. Our results indicate that OFFESEL outperforms all supervised and unsupervised feature selection algorithms in terms of classification accuracy. Specifically, OFFESEL preserves the accuracy level of the supervised FIRES algorithm, which proved more accurate than ABFS in our experiments, while maintaining the accuracy level achieved by the unsupervised Max Variance algorithm. Moreover, OFFESEL requires even less computation time than Max Variance and shows high stability on stationary datasets. Overall, our study demonstrates the potential benefits of using unlabeled data for feature ranking and selection in dynamic data streams.

Discursive Patinas: Anchoring Discussions in Data Visualizations.

This paper presents discursive patinas, a technique to visualize discussions onto data visualizations, inspired by how people leave traces in the physical world. While data visualizations are widely discussed in online communities and social media, comments tend to be displayed separately from the visualization and we lack ways to relate these discussions back to the content of the visualization, e.g., to situate comments, explain visual patterns, or question assumptions. In our visualization annotation interface, users can designate areas within the visualization. Discursive patinas are made of overlaid visual marks (anchors), attached to textual comments with category labels, likes, and replies. By coloring and styling the anchors, a meta visualization emerges, showing what and where people comment and annotate the visualization. These patinas show regions of heavy discussions, recent commenting activity, and the distribution of questions, suggestions, or personal stories. We ran workshops with 90 students, domain experts, and visualization researchers to study how people use anchors to discuss visualizations and how patinas influence people's understanding of the discussion. Our results show that discursive patinas improve the ability to navigate discussions and guide people to comments that help understand, contextualize, or scrutinize the visualization. We discuss the potential of anchors and patinas to support discursive engagements, including critical readings of visualizations, design feedback, and feminist approaches to data visualization.

Preferred labels and language to improve communication about lesions at low risk of progressing to cancer: qualitative interviews with patients and physicians

ObjectivesWe explored how to improve communication about low-risk lesions including labels, language and other strategies.DesignQualitative description and thematic analysis to examine the transcripts of telephone interviews with patients who had...

Leveraging Partial Labels for Cervical Lesion Classification via a Multilabel Approach

Leveraging Partial Labels for Cervical Lesion Classification via a Multilabel Approach

MMSeg: A novel multi-task learning framework for class imbalance and label scarcity in medical image segmentation

MMSeg: A novel multi-task learning framework for class imbalance and label scarcity in medical image segmentation

Exploration of designing an automatic classifier for questions containing code snippets-A case study of Oracle SQL certification exam questions.

This study uses the Oracle SQL certification exam questions to explore the design of automatic classifiers for exam questions containing code snippets. SQL's question classification assigns a class label in the exam topics to a question. With this classification, questions can be selected from the test bank according to the testing scope to assemble a more suitable test paper. Classifying questions containing code snippets is more challenging than classifying questions with general text descriptions. In this study, we use factorial experiments to identify the effects of the factors of the feature representation scheme and the machine learning method on the performance of the question classifiers. Our experiment results showed the classifier with the TF-IDF scheme and Logistics Regression model performed best in the weighted macro-average AUC and F1 performance indices. The classifier with TF-IDF and Support Vector Machine performed best in weighted macro-average Precision. Moreover, the feature representation scheme was the main factor affecting the classifier's performance, followed by the machine learning method, over all the performance indices.

A General Framework for Comparing Embedding Visualizations Across Class-Label Hierarchies.

Projecting high-dimensional vectors into two dimensions for visualization, known as embedding visualization, facilitates perceptual reasoning and interpretation. Comparing multiple embedding visualizations drives decision-making in many domains, but traditional comparison methods are limited by a reliance on direct point correspondences. This requirement precludes comparisons without point correspondences, such as two different datasets of annotated images, and fails to capture meaningful higher-level relationships among point groups. To address these shortcomings, we propose a general framework for comparing embedding visualizations based on shared class labels rather than individual points. Our approach partitions points into regions corresponding to three key class concepts-confusion, neighborhood, and relative size-to characterize intra- and inter-class relationships. Informed by a preliminary user study, we implemented our framework using perceptual neighborhood graphs to defne these regions and introduced metrics to quantify each concept. We demonstrate the generality of our framework with usage scenarios from machine learning and single-cell biology, highlighting our metrics' ability to draw insightful comparisons across label hierarchies. To assess the effectiveness of our approach, we conducted an evaluation study with fve machine learning researchers and six single-cell biologists using an interactive and scalable prototype built with Python, JavaScript, and Rust. Our metrics enable more structured comparisons through visual guidance and increased participants' confdence in their fndings.

IMPLEMENTATION OF BALANCING DATA METHOD USING SMOTETOMEK IN DIABETES CLASSIFICATION USING XGBOOST

In this research, XGBoost algorithm and the SMOTETomek approach are employed with the objective of enhancing the accuracy of diabetes classification. The study utilises 2,000 patient data points, comprising demographic and medical information, sourced from Kaggle. The dataset employed in this study comprises a number of variables, including pregnancies, glucose levels, blood pressure, skin thickness, insulin levels, Body Mass Index (BMI), diabetes pedigree function, age, and an outcome variable. The latter is a binary classification label, taking on the values 0 and 1. A value of 0 indicates that the patient is not affected by diabetes, whereas a value of 1 indicates that the patient has diabetes. Diabetes represents a significant public health concern in Indonesia. A significant challenge in this study was the imbalanced nature of the dataset, which included a disproportionate number of non-diabetic samples relative to diabetic samples. To address this class imbalance, the researchers employed the SMOTETomek method. SMOTETomek integrates the SMOTE (Synthetic Minority Over-sampling Technique) and Tomek links algorithms to oversample the minority class and remove borderline samples, thereby balancing the class distributions. The SMOTETomek method achieved higher accuracy (95.01%) than SMOTE and the original data (both 92.13%), highlighting the benefits of combining SMOTE with Tomek Links for XGBoost. During testing, SMOTETomek slightly reduced the minority class accuracy (0.97 vs. 0.99 for SMOTE and original data) but maintained strong F1-score and precision, indicating effective handling of data imbalance despite minor trade-offs.

Weakly Supervised Instance Segmentation in Aerial Images via Comprehensive Spatial Adaptation

Weakly supervised instance segmentation (WSIS) only employs image-level supervision to identify instance class labels and create segmentation masks, drawing increasing attention. Currently, existing WSIS methods primarily focus on activating the most discriminative regions and then inferring the entire instance by analyzing inter-pixel relationships within those regions. However, these identification regions are typically concentrated in limited but critical regions or are mistakenly activated in the background region, making it challenging to address scale variations among instances. Furthermore, different aerial instances often appear in close proximity, resulting in the merging of multiple instances of the same class. To tackle these challenges, we propose a comprehensive approach called Comprehensive Spatial Adaptation Segmentation (CSASeg). Specifically, the self-adaptive spatial-aware enhancement network (SSE) identifies extensive regions by analyzing spatial consistency within the class semantic map. Then, we develop a multi-level projection field (MPF) module to significantly enhance instance-level discrimination through deep-to-shallow residual estimation. Additionally, a foreground enhancement module is incorporated into SSE to reduce background noise while enhancing foreground details, significantly increasing the effectiveness of instance analysis. Extensive experiments conduct on three challenging datasets, iSAID, NWPU VHR-10.v2, and SSDD, demonstrate the competitiveness of our proposed approach.

ResNet50-3Cur-HGCN: a novel multimodal hybrid curvature space approach to bearing fault diagnosis

Abstract Bearings are widely used in various fields due to their advantages of high load-bearing capacity, low friction loss, and minimal vibration and impact. In response to the current issues in the process of fault signal identification using graph neural networks, such as the insufficient expression of node data types, the inadequate exploration of feature information carried by the graph structure, and the singularity of mathematical space operations on graph data, this paper proposes a ResNet50-HGCN model based on the combination of Residual Network-50 (ResNet50) and Heterogeneous Graph Convolutional Neural Network (HGCN). The model is trained in a threefold mixed curvature (3 Curvature, 3Cur) space environment for semi-supervised learning, aiming to achieve accurate classification of bearing fault signals. Specifically, first, the time-domain signals and the components of the time-frequency map obtained from the Wavelet Synchrosqueezing Transform (WSST) are used as the bimodal node information for HGCN. Then, the graph network is merged in the 3Cur space for training and weighted validation, obtaining the predicted category labels of the model under the 3Cur configuration space. Finally, experimental data analysis is conducted, sparse confusion matrices for predicted categories are drawn, and four types of accuracy-related evaluation metrics are calculated. The experimental results show that the proposed ResNet50-3Cur-HGCN classification model outperforms other models in the experiment, achieving an accuracy of 97.71%, which verifies the advantage of the method in terms of accuracy and effectiveness. It also provides a good methodological reference for bearing fault diagnosis approaches centered on graph neural networks.

Deep superpixel generation and clustering for weakly supervised segmentation of brain tumors in MR images

PurposeTraining machine learning models to segment tumors and other anomalies in medical images is an important step for developing diagnostic tools but generally requires manually annotated ground truth segmentations, which necessitates significant time and resources. We aim to develop a pipeline that can be trained using readily accessible binary image-level classification labels, to effectively segment regions of interest without requiring ground truth annotations.MethodsThis work proposes the use of a deep superpixel generation model and a deep superpixel clustering model trained simultaneously to output weakly supervised brain tumor segmentations. The superpixel generation model’s output is selected and clustered together by the superpixel clustering model. Additionally, we train a classifier using binary image-level labels (i.e., labels indicating whether an image contains a tumor), which is used to guide the training by localizing undersegmented seeds as a loss term. The proposed simultaneous use of superpixel generation and clustering models, and the guided localization approach allow for the output weakly supervised tumor segmentations to capture contextual information that is propagated to both models during training, resulting in superpixels that specifically contour the tumors. We evaluate the performance of the pipeline using Dice coefficient and 95% Hausdorff distance (HD95) and compare the performance to state-of-the-art baselines. These baselines include the state-of-the-art weakly supervised segmentation method using both seeds and superpixels (CAM-S), and the Segment Anything Model (SAM).ResultsWe used 2D slices of magnetic resonance brain scans from the Multimodal Brain Tumor Segmentation Challenge (BraTS) 2020 dataset and labels indicating the presence of tumors to train and evaluate the pipeline. On an external test cohort from the BraTS 2023 dataset, our method achieved a mean Dice coefficient of 0.745 and a mean HD95 of 20.8, outperforming all baselines, including CAM-S and SAM, which resulted in mean Dice coefficients of 0.646 and 0.641, and mean HD95 of 21.2 and 27.3, respectively.ConclusionThe proposed combination of deep superpixel generation, deep superpixel clustering, and the incorporation of undersegmented seeds as a loss term improves weakly supervised segmentation.

Acquiring complex concepts through classification versus observation

We report six experiments that examine how two essential components of a category-learning paradigm, training and feedback, can be manipulated to maximize learning and transfer of real-world, complex concepts. Some subjects learned through classification and were asked to classify hypothetical experiment scenarios as either true or non-true experiments; others learned through observation, wherein these same scenarios were presented along with the corresponding category label. Additionally, some subjects were presented correct-answer feedback (the category label), whereas others were presented explanation feedback (the correct answer and a detailed explanation). For classification training, this feedback was presented after each classification judgment; for observation training this feedback was presented simultaneously with the hypothetical experiment. After the learning phase, subjects completed a posttest that included one task that involved classifying novel hypothetical scenarios and another task comprising multiple-choice questions about novel scenarios, in which subjects had to specify the issue with the scenario or indicate how it could be fixed. The posttest either occurred immediately after the learning phase (Experiments 1–2), 10 min later (Experiments 3–4), two days later (Experiment 5), or one week later (Experiment 6). Explanation feedback generally led to better learning and transfer than correct-answer feedback. However, although subjects showed clear evidence of learning and transfer, posttest performance did not differ between classification and observation training. Critically, various learning theories and principles (e.g., retrieval practice, generation, active learning) predict a classification advantage. Our results thus call into question the extent to which such theories and principles extend to the transfer of learning.

Comparative Analysis of Machine Learning Models to Predict Common Vulnerabilities and Exposure

Predicting Common Vulnerabilities and Exposures (CVE) is a challenging task due to the increasing complexity of cyberattacks and the vast amount of threat data available. Effective prediction models are crucial for enabling cybersecurity teams to respond quickly and prevent potential exploits. This study aims to provide a comparative analysis of machine learning techniques for CVE prediction to enhance proactive vulnerability management and strengthening cybersecurity practices. The supervised machine learning model which is Gaussian Naive Bayes and unsupervised machine learning models that utilize clustering algorithms which are K-means and DBSCAN were employed for the predictive modelling. The performance of these models was compared using performance metrics such as accuracy, precision, recall, and F1-score. Among these models, the Gaussian Naive Bayes achieved an accuracy rate of 99.79%, and outperformed the clustering-based machine learning models in effectively determining the class labels or results of the data it was trained on or tested against. The outcome of this study will provide a proof of concept to Cybersecurity Malaysia, offering insights into the CVE model.

Leveraging Deep Learning for Robust Structural Damage Detection and Classification: A Transfer Learning Approach via CNN

Transfer learning techniques for structural health monitoring in bridge-type structures are investigated, focusing on model generalizability and domain adaptation challenges. Finite element models of bridge-type structures with varying geometry were simulated using the OpenSeesPy platform. Different levels of damage states were introduced at the midspans of these models, and Gaussian-based load time histories were applied at mid-span for dynamic time-history analysis to calculate acceleration data. Then, this acceleration time-history series was transformed into grayscale images, serving as inputs for a Convolutional Neural Network developed to detect and classify structural damage states. Initially, it was trained and tested on datasets derived from a Single-Source Domain structure, achieving perfect accuracy (1.0) in a ten-label multi-class classification task. However, this accuracy significantly decreased when the model was sequentially tested on structures with different geometry without retraining. To address this challenge, it is proposed that transfer learning be employed via feature extraction and joint training. The model showed a reduction in accuracy percentage when adapting from a Single-Source Domain to Multiple-Target Domains, revealing potential issues with non-homogeneous data distribution and catastrophic forgetting. Conversely, joint training, which involves training on all datasets except the specific Target Domain, generated a generalized network that effectively mitigated these issues and maintained high accuracy in predicting unseen class labels. This study highlights the integration of simulation data into the Deep Learning-based SHM framework, demonstrating that a generalized model created via Joint Learning utilizing FEM can potentially reduce the consequences of modeling errors and operational uncertainties unavoidable in real-world applications.

Building robust malware detection through conditional Generative Adversarial Network-based data augmentation

Malware detection is essential in cybersecurity, yet its accuracy is often compromised by class imbalance and limited labeled data. This study leverages conditional Generative Adversarial Networks (cGANs) to generate synthetic malware samples, addressing these challenges by augmenting the minority class. The cGAN model generates realistic malware samples conditioned on class labels, balancing the dataset without altering the benign class. Applied to the CICMalDroid2020 dataset, the augmented data is used to train a LightGBM model, leading to improved detection accuracy, particularly for underrepresented malware classes. The results demonstrate the efficacy of cGANs as a robust data augmentation tool, enhancing the performance and reliability of machine learning-based malware detection systems.

Pseudo Multi-Modal Approach to LiDAR Semantic Segmentation.

To improve the accuracy and reliability of LiDAR semantic segmentation, previous studies have introduced multi-modal approaches that utilize additional modalities, such as 2D RGB images, to provide complementary information. However, these methods increase the cost of data collection, sensor hardware requirements, power consumption, and computational complexity. We observed that multi-modal approaches improve the semantic alignment of 3D representations. Motivated by this observation, we propose a pseudo multi-modal approach. To this end, we introduce a novel class-label-driven artificial 2D image construction method. By leveraging the close semantic alignment between image and text features of vision-language models, artificial 2D images are synthesized by arranging LiDAR class label text features. During training, the semantic information encoded in the artificial 2D images enriches the 3D features through knowledge distillation. The proposed method significantly reduces the burden of training data collection and facilitates more effective learning of semantic relationships in the 3D backbone network. Extensive experiments on two benchmark datasets demonstrate that the proposed method improves performance by 2.2-3.5 mIoU over the baseline using only LiDAR data, achieving performance comparable to that of real multi-modal approaches.