Search Results (79)

Reliable in-situ defect classification is essential for ensuring the mechanical performance of parts produced by material extrusion. However, machine learning and deep learning models often suffer from poor generalization because training datasets typically consist of many correlated signal segments generated from only a limited number of independent print jobs. To address this problem, we analyze how job-specific characteristics are distributed across signal components and find that they are concentrated primarily in the signal level rather than in the residual component. Motivated by this observation, we propose Level-Aware Residual Mixup(LARM), a data augmentation method that separately interpolates the level and residual components of sensor signals. LARM preserves realistic job-level characteristics by restricting level interpolation to values observed in real print jobs while allowing flexible mixing of residual components within the same defect class. We evaluate LARM against representative augmentation methods across diverse classification models under a realistic job-level leave-one-group-out cross-validation protocol. Experimental results demonstrate that LARM achieves better generalization than both training without augmentation and representative augmentation methods. Full article

Multimodal behavioral sensing may support depression screening, but evaluation on small clinical-interview datasets is particularly vulnerable to data leakage and model-selection bias. We present a leakage-audited trimodal framework evaluated on DAIC-WOZ ($n=180$, PHQ-8 $\ge 10$), combining SBERT text embeddings, OpenFace facial-behavior descriptors, and COVAREP prosodic features. Participant-level partitioning is performed before augmentation, while decision thresholds and neural-model checkpoints are selected exclusively from internal validation data. A controlled five-seed experiment showed that a deliberately leaky full-pool MixUp construction, in which a retained development sample could include a held-out participant as its second parent, was associated with a 27–33 percentage-point increase in Macro-F1 across four fusion configurations. Under the participant-level leakage-free 5-fold protocol, trimodal late fusion achieved a Macro-F1 of $0.532\pm 0.041$, compared with $0.446\pm 0.021$ for Text+Imaging late fusion. Paired participant-level correctness outcomes also favored trimodal fusion (McNemar ${\chi }^2=6.618$, $p=0.010$), consistent with improved paired classification when the audio modality was included under the leakage-free protocol. On 86 participant-disjoint E-DAIC sessions, using a consistent PHQ-8-based outcome definition (PHQ-8 $\ge 10$), trimodal late fusion achieved Macro-F1 = 0.621 and AUC = 0.676; this experiment is interpreted as within-family generalization rather than independent cross-corpus validation. Overall, the results show that leakage control can substantially alter both absolute performance and comparative conclusions, and that multimodal gains should be established through participant-level evaluation, modality-specific analysis, and reproducible model-selection procedures. Full article

Background/Objectives: Class imbalance is one of the major challenges in automated skin lesion classification since the number of categories of malignant and clinically significant skin lesions is normally less than the benign ones. However, due to this imbalance, deep convolutional neural networks (CNNs) tend to overlook minority classes and fail to recognize them with an acceptable accuracy, which leads to a decrease in diagnostic reliability. A wide range of imbalance mitigation techniques has been suggested, but their effectiveness is found to differ significantly depending on CNN architecture, and detailed comparative studies of these techniques for a consistent experimental setup are still limited. Methods: This study proposes a comprehensive benchmarking framework that tests sixteen class imbalance mitigation methods by applying them to six pretrained CNN architectures—EfficientNet-B0, EfficientNet-B3, ResNet50, DenseNet121, InceptionV3 and MobileNetV2—on the official ISIC 2019 skin lesion dataset. The tested techniques are conventional resampling techniques, synthetic sample generation techniques, algorithm-level learning techniques, data augmentation techniques, and hybrid techniques. The dataset was partition into a separate training set and testing set, and stratified cross-validation was only conducted on the training set to ensure the study was fair and reproducible. Both models have been optimized with the same optimizer, learning rate, batch size, epochs and preprocessing pipeline. The performance of the models was evaluated by computing the accuracy, precision, recall and F1-score. Results: The experimental results show that the effect of class imbalance mitigation is very specific to the underlying CNN architecture. The traditional undersampling and oversampling methods yielded only moderate improvements, while feature space and hybrid methods yielded more consistent results. When coupled with EfficientNet-B3, Balanced MixUp improved the overall performance of the model by achieving an accuracy of 92.39%, an increase in precision of 93.3%, a recall of 91.36%, and an F1-score of 92.33%. However, some architectures such as ResNet50 performed better with iterative learning techniques, such as Cumulative Learning and Yielding Multi-Fold Training, which suggests that there is a diversity in how different network architectures react to imbalance mitigation methods. Conclusions: This paper highlights the importance of selecting appropriate technique–architecture combinations for addressing long-tailed data distributions in medical imaging. The proposed benchmarking framework provides valuable insights for developing robust and reliable deep learning systems for skin lesion classification and other medical imaging tasks affected by severe class imbalance. Full article

Graph Neural Networks (GNNs) have achieved promising performance in various graph learning tasks. Real-world graph learning often suffers from severe label scarcity, which hinders GNNs’ feature and topology learning and causes poor generalization. Node mixup has therefore been shown to be an effective graph data augmentation paradigm for enhancing GNNs’ performance in label-scarce graph learning scenarios. However, mainstream node mixup methods adopt fixed or randomly sampled mixing ratios while neglecting inherent correlations between node pairs, thereby producing less informative or noisy augmented samples. To address the issue, we propose a novel semantic-aware adaptive node mixup method, SA-Mixup, for node classification tasks. Specifically, SA-Mixup first leverages high-confidence pseudo-labeling to enrich supervisory signals by incorporating reliable unlabeled nodes into the training set, and then it selects node pairs with the same label for node mixup. Furthermore, an adaptive node mixup mechanism is developed to dynamically learn optimal mixing ratios according to semantic relation and predictive uncertainty for each node pair. In addition, a similarity-guided neighbor connection strategy is introduced to select candidate neighbors for generated virtual nodes and establish reasonable topological connections. The experimental results show that the proposed method achieves strong performance when combined with different GNN backbones. The ablation results further show that both the adaptive node mixup mechanism and the similarity-guided neighbor connection contribute to the performance improvements. Full article

The design of cushioning packaging for flat-screen television (TV) products relies heavily on repeated simulations, resulting in high development costs and low design efficiency. In this study, we propose a hybrid framework integrating finite element (FE) simulation, data augmentation and interpretable machine learning (ML) for rapid peak acceleration prediction and optimization of TV cushioning packaging. First, a total of 216 FE drop-impact simulation samples of TV cushioning packaging systems were generated using ANSYS Workbench, covering TV dimensions, liner type, liner density, liner thickness, drop height and peak acceleration. Mixup-based data augmentation and Bayesian optimization were then employed to develop and tune six ML models. All ML models trained on the original dataset achieved coefficients of determination (R²) ranging from 0.797 to 0.990. The Mixup-augmented XGBoost model achieved the best prediction performance, yielding R² values of 0.998 and 0.983 for the training and testing datasets, respectively. SHAP analysis revealed that liner material type, liner density and liner thickness were the dominant factors affecting the protective performance of TV cushioning packaging. In addition, a web-based platform was developed based on the proposed FE–ML strategy to support the design exploration of feasible schemes for new TV products. The predictive capability of the proposed FE-ML framework was further evaluated using 22 independent cushioning packaging schemes, achieving an R² of 0.926 and an average prediction error of 4.490 g. These results suggest that the proposed workflow can support the performance evaluation and optimization of TV cushioning packaging. Full article

The positioning and clamping process in aircraft assembly exhibits pronounced long-term temporal correlations and intense human–machine interactions. Consequently, assembly quality depends heavily on operator compliance and consistency. Capturing long-term, strongly correlated features in complex industrial environments remains a significant challenge. To overcome this, this study proposes a Long-Term Strongly Associated Action Recognition Network (LTSA-Net) tailored for aircraft assembly positioning and clamping tasks. Based on the C3D backbone, the model first incorporates the SimAM attention mechanism and BN modules to significantly enhance focus on critical spatiotemporal features. To address the challenge of capturing long-term temporal dependencies, LTSFEM is designed to extract global temporal information accurately. Furthermore, to balance structural lightweight design with real-time inference requirements, the CWSTB module is integrated to achieve substantial parameter compression. In addition, a dedicated aircraft assembly positioning and clamping dataset was constructed, and a robust training framework was established using the AdamW optimizer and Mixup data augmentation. Experimental results demonstrate that LTSA-Net achieves a recognition accuracy of 98.82% on the LTSA-Dataset, with a per-frame inference time of 42 ms, successfully meeting the dual requirements of high precision and real-time performance in industrial scenarios, and providing a practical technical solution for intelligent monitoring of aircraft assembly processes. Full article

High-precision identification of acoustic convergence zones (CZs) and acoustic shadow zones (SZs) is a core prerequisite for deep-sea sonar performance prediction and long-range underwater target detection. However, in data-scarce marine environments, traditional acoustic identification methods suffer from high environmental sensitivity and significant computational costs, while pure data-driven deep learning methods face dilemmas such as a lack of physical consistency and poor generalization on small samples. To address these issues, a three-level cascaded recognition framework based on physics-prior-augmented deep learning is proposed in this paper, enabling accurate segmentation of CZs and intelligent classification of sound field types under data-scarce scenarios. In this framework, physical acoustic principles are incorporated exclusively as priors through a training dataset generated by a Gaussian beam acoustic propagation code (Bellhop) and through hand-crafted geometric features derived post hoc from the initial segmentation outputs. Taking a typical deep-sea area in the Northwest Pacific Ocean as the research object, a hybrid dataset comprising 5000 simulated transmission loss images and 500 simulated images from a geographically distinct sea area is constructed. The sound field is categorized into four types: strong convergence, usable convergence, weak convergence, and shadow zone. In the first stage, the ResNet-34 backbone is improved by integrating deformable convolution and a global statistical feature module, which, combined with a joint loss function, achieves high-precision pixel-level segmentation of CZs and SZs, with the regional gray contrast reaching 86.9%. In the second stage, a customized dual-channel VGG16 architecture is designed to fuse the extracted geometric priors and visual features, achieving a sound field classification accuracy of 89.91%. In the third stage, a hybrid data augmentation technique combining Mixup and convolutional autoencoder is adopted alongside a transfer learning strategy to mitigate the data scarcity under cross-domain conditions, boosting the small-sample classification accuracy to 84.45%. The experimental results demonstrate that the models in each stage of the proposed framework significantly outperform traditional methods and baseline networks. This study provides a novel methodology and technical support for intelligent sound field identification in data-scarce marine environments. Finally, the core contributions and current limitations are summarized, and future research directions, such as constructing a dynamic hydrological parameter feedback mechanism and identifying three-dimensional complex sound fields, are prospected. Full article

Blood cell detection is an important fundamental step in assisting the diagnosis of hematological diseases. However, existing deep learning-based detection methods usually rely on large amounts of high-quality manually annotated data, while medical image annotation is costly, time-consuming, and requires professional expertise. To address the above problems, a consistency regularization-based semi-supervised blood cell detection method, CRS-YOLOv5-ALT, is proposed, which is built upon the YOLOv5-ALT framework. Specifically, an initial detection model is first trained using a small number of labeled samples. Then, the current model is used to predict unlabeled samples, and candidate detection results are filtered through a dual-threshold pseudo-label filtering strategy by combining confidence and prediction entropy thresholds to generate relatively reliable pseudo-labels. On this basis, a consistency regularization strategy is introduced, in which consistency constraints between original unlabeled images and perturbed augmented images are constructed to guide the model to maintain stable prediction results under input perturbations. Meanwhile, a Mixup data augmentation strategy suitable for object detection tasks is introduced to enrich the training sample distribution and further improve the generalization capability of the model. The experimental results on the BCCD dataset show that, when only 5% labeled data are used, CRS-YOLOv5-ALT achieves an mAP of 59.73 ± 0.16%, an mAP50 of 91.97 ± 0.13%, and an mAP75 of 65.82 ± 0.19%, outperforming representative semi-supervised methods such as STAC, Instant Teacher, and CSD. Additional analyses of repeated runs, pseudo-label evolution, filtering acceptance rate, confidence- and entropy-based pseudo-label reliability indicators, training convergence, and ablation experiments provide further quantitative support for the proposed method. In addition, the single-run evaluation on the TXL-PBC dataset provides supplementary evidence for the cross-dataset applicability of CRS-YOLOv5-ALT under the current experimental setting. Overall, CRS-YOLOv5-ALT can effectively utilize unlabeled data under low-annotation conditions and provides a feasible semi-supervised learning strategy for reducing the annotation cost of blood cell images. Full article

Mixup is a data augmentation technique that improves prediction accuracy in classification tasks by combining representations of training samples, which makes it particularly effective in settings with limited data and during fine-tuning for downstream tasks. However, representations generated by mixup may appear unnatural, which can negatively affect fine-tuning performance. To address this limitation, we propose a vision transformer (ViT)-aware variant of mixup strategies, Dimensional Swap Mix (DiSMix). DiSMix divides a representation vector into two segments corresponding to subspaces of the original feature space and generates new representations by swapping one segment with that from another sample and concatenating the segments. This allows part of the original representation to remain unchanged, enabling the model to learn from partially preserved features. We evaluate DiSMix by applying several mixup-based methods to fine-tune ViTs on the VTAB-1k benchmark. The findings show that DiSMix improves accuracy on the VTAB-1k Natural split, reaching 80.0%, compared with conventional mixup methods. This suggests that DiSMix is an effective alternative for representation-level data augmentation in fine-tuning scenarios. Full article

To address the problem in predicting target prices for ship equipment orders where small sample sizes, high feature dimensions, and strong business constraints lead traditional models to overfit and have insufficient generalization ability, a method combining GAN and NVP-D for small-sample data augmentation and price prediction is proposed. This method integrates the advantages of adversarial training in Generative Adversarial Networks (GAN) with the explicit density estimation and stable training characteristics of Normalizing Flow NVP-D. By using dual-weight collaborative optimization of the objective function, it alleviates gradient vanishing and mode collapse, generating high-quality virtual samples that closely follow the real data distribution. Redundant features are removed using Boruta-Lasso joint feature selection to reduce model complexity. CatBoost is employed as the prediction model to complete price estimation. Experiments were conducted on a ship equipment dataset with 33 original samples and 24 features, strictly following the standard procedure of augmentation only within the training set and 5-fold cross-validation. Compared with NVP, NVP-G, MAF, traditional GAN, and Mixup methods, the results show that the proposed integrated model achieves optimal performance when augmenting 400 samples, with an RMSE of 0.0675, MAE of 0.0510, and R² of 0.9228. After feature selection, prediction accuracy further improves, with RMSE decreasing to 0.0615 and R² increasing to 0.9341. Limited by the scale of the original samples, the statistical robustness and cross-dataset generalization capability of this method still need validation with larger datasets. However, under the current small-sample constraints, it can effectively alleviate modeling bottlenecks and provide high-precision support for equipment procurement argumentation, budget preparation, and cost control in stages. Full article

Unsupervised medical image translation remains challenging because model development often relies on unpaired training, whereas reliable evaluation requires well-matched reference images. PD-weighted and T2-weighted brain MR images provide a useful testbed for this problem because they are closely matched anatomically while still exhibiting distinct contrast characteristics. Existing methods often align only high-level features, overlooking low-level texture details that are important for structural fidelity. In this work, we propose the Multi-Scale Feature Regularization and Patch Mixup (MSFRPM) framework based on an encoder–decoder architecture. It aligns cross-domain features across multiple scales to preserve local details and employs a patch-based mixup strategy to augment training data. The framework was evaluated using an unsupervised learning protocol with strict data partitioning. Experimental results demonstrate that MSFRPM achieves strong performance relative to eight state-of-the-art methods. Our approach achieved improvements in MAE (6.26 ± 0.86), PSNR (23.53 ± 0.92), SSIM (0.83 ± 0.03), and GMSD (0.100 ± 0.010). Qualitative assessments confirmed improved structural fidelity, and t-SNE visualization validated enhanced cross-domain feature alignment. Overall, MSFRPM provides a useful approach for unsupervised PD-to-T2 image translation under the current experimental setting. Full article

Reliable photovoltaic (PV) power forecasting based on deep learning typically requires large historical datasets to capture the high temporal and spatial variability of solar irradiance. However, in many real-world applications, data availability is limited to short observation periods, hindering the effective training of deep learning models. This paper investigates how sky image data augmentation techniques can improve the generalization capability of Convolutional Neural Networks (CNNs) trained under data scarcity. Three augmentation-based oversampling methods—SMOTE, Mixup-kNN, and Mixup-RP—are evaluated, along with two novel hybrid strategies that combine these methods in parallel and series configurations. The proposed framework is validated on two distinct PV power nowcasting case studies, in which the original sky image training datasets span less than one month. Experimental results show average performance improvements of up to 50% on external testing data when training the CNN on the augmented datasets compared to the original base datasets, demonstrating that accurate PV power nowcasting is feasible even under data-scarce conditions typical of newly installed PV systems, and highlighting the potential of data-efficient learning approaches for renewable energy applications. Full article

Software effort estimation (SEE) serves as a cornerstone of effective software project management, and case-based reasoning (CBR) stands out as one of the most extensively adopted approaches within this domain. Nevertheless, CBR-based SEE models are still plagued by two critical challenges: conventional case retrieval mechanisms lack the ability to differentiate the relative importance of various features, and data scarcity remains a persistent bottleneck. Both issues significantly compromise the estimation accuracy and interpretability of the models. To address these limitations, we propose a SHAP–Mixup synergistic framework that enhances both feature-aware similarity learning and data distribution modeling. Specifically, we introduce (1) a stability-aware SHAP-weighted similarity metric that integrates both the magnitude and variance of feature contributions to improve retrieval robustness, and (2) a density-aware Mixup augmentation strategy that generates synthetic samples guided by local data manifold structure rather than random interpolation. Experimental results on seven benchmark datasets demonstrate that the proposed method reduces MAE and MSE by up to 20.2% on average compared to baseline CBR models, while consistently improving Pred(0.25). Furthermore, by enhancing model interpretability, the proposed method equips project managers with actionable insights into the key drivers of software effort, thereby facilitating more informed and efficient resource allocation. Building on these findings, this study provides a novel and effective pathway for developing SEE models that are more accurate, robust, and transparent. Full article

Background: Thyroid nodule classification on ultrasound remains challenging due to limited labeled data and marked class imbalance. This study proposes an integrated deep learning framework combining YOLO-based region-of-interest detection with an enhanced ResNet18 classifier. Methods: A total of 522 thyroid ultrasound images from 522 patients examined between July 2020 and June 2024 were included. The dataset comprised 467 images for training (399 benign, 68 malignant), 41 for independent testing (19 benign, 22 malignant), and 14 for internal validation (4 benign, 10 malignant). An external validation set of 36 images (22 benign, 14 malignant) was collected from online sources. ResNet18 with a convolutional block attention module was used to enhance feature extraction. To address small sample size and class imbalance, the training pipeline incorporated focal loss, weighted random sampling, mixup augmentation, cosine annealing learning rate scheduling, and a 5-fold cross-validation ensemble. Results: The ensemble model achieved 85.4% accuracy (95% CI: 74.5–96.2%), 86.4% sensitivity (95% CI: 72.0–100%), and 84.2% specificity (95% CI: 67.8–100%) on the independent test set. Internal validation yielded 85.7% accuracy, 90.0% sensitivity, and 75.0% specificity, while external validation demonstrated 77.8% accuracy, 78.6% sensitivity, and 77.3% specificity. These findings suggest that advanced regularization combined with ensemble learning improves generalizability despite limited data. Conclusions: This study demonstrates that a lightweight ResNet18 architecture with strategic optimization outperforms deeper networks on small medical datasets. The proposed framework demonstrated good diagnostic performance across multiple validation cohorts, offering a promising computer-aided diagnosis tool for thyroid nodule assessment. Full article

Accurate and efficient modeling of AlGaN/GaN HEMTs is essential for the design of next-generation power electronics. This study introduces a hybrid Auxiliary Classifier Generative Adversarial Network (ACGAN)–mixup data augmentation framework to enhance deep neural network application in AlGaN/GaN high-electron-mobility transistor modeling with limited data. Based on only 20 distinctive devices, ACGAN uses technology computer-aided design (TCAD)-calibrated data to generate high-quality synthetic drain current ($I_{ds}$) under various electronic bias conditions. The quality of the generated data is validated via Jensen–Shannon divergence with an average of 0.0341. A one-dimensional convolutional neural network (1D-CNN) predictive model is trained on augmented data and achieves stable convergence, with a mean absolute error of 0.002 A/mm for the off-state $I_{ds}$ and 0.052 A/mm for the linear region. It also shows improved robustness over the model trained on original non-augmented data. The proposed approach offers a low-cost alternative to resource-intensive TCAD simulations, enabling accurate device modeling with limited data. Full article