Search Results (786)

Cross-Domain Few-Shot Scene Classification (CDFSSC) aims to transfer knowledge from a source domain to a target domain for few-shot classification tasks, and is essential for remote sensing applications involving diverse platforms and dynamic environments. However, distribution discrepancies and category misalignment across domains often introduce high predictive uncertainty, significantly degrading model performance. To address these challenges, an uncertainty-aware cross-domain (UACD) framework is proposed to enhance model reliability by systematically mining uncertainty-related information. Specifically, in the cross-domain training process, a feature-decision consistency regularization (FDCR) structure is designed to stabilize cross-domain training by enforcing consistency at both feature and decision levels. Furthermore, an uncertainty-aware knowledge mining (UKM) policy is introduced to effectively exploit high-uncertainty target samples, mitigating the negative impact of unreliable pseudo-labels and improving representation learning. In the few-shot adaptation stage, an uncertainty-aware predictor is developed to enhance adaptability and decision-making in target tasks. Extensive experiments on 12 cross-domain scenarios demonstrate that the proposed UACD framework consistently achieves superior or competitive performance, with strong robustness and generalization capability across diverse CDFSSC tasks. Full article

The running reliability of rolling bearings depends on the effective lubrication state, and poor lubrication will induce abnormal vibration. Therefore, vibration-based fault diagnosis is an important means to evaluate the health of bearings through vibration characteristics. However, the lack of fault samples in actual working conditions seriously restricts the generalization ability and accuracy of an intelligent diagnosis model. A novel few-shot diagnosis method integrating multi-layer feature fusion and adaptive similarity measurement is proposed. This method adopts a meta-learning framework to simulate sample scarcity through numerous N-way K-shot diagnostic tasks. An efficient feature extractor with a cross-task feature stitching mechanism is designed to fuse features from support and query sets. To overcome the limitation of fixed-distance metrics in existing meta-learners, a learnable similarity scheduler adaptively generates optimal pseudo-distance functions. In particular, a multi-layer feature fusion strategy is introduced to compute adaptive similarities at multiple network depths, which significantly enhances feature robustness against operational variations. Experimental results demonstrate the method achieves stable diagnostic accuracy above 90% under extremely few-shot conditions and maintains over 90% accuracy when transferring from laboratory-simulated faults to natural operational faults, validating its strong potential for practical industrial applications where annotated fault data is scarce. Full article

Traditional malware classification approaches primarily address fixed sets of well-studied malware types and therefore struggle to accommodate the continual emergence of novel or previously unseen malware strains. While visualization-based strategies have shown promise in few-shot malware classification, existing methods often produce representations with limited semantic richness. In parallel, few-shot learning models frequently converge with suboptimal solutions, limiting their ability to generalize effectively to new classes. To address these challenges, we propose MetaWave, a unified framework that jointly optimizes both data representation and model learning for few-shot malware classification. Rather than treating feature representation and learning strategy as largely independent stages, MetaWave is formulated as an explicit representation–adaptation integration framework that combines multi-view malware encoding with meta-learning-based optimization. At the data level, we propose a Wavelet Transform-based Malware Representation method that leverages multi-scale frequency analysis and complementary views to generate semantically enriched representations. At the model level, we adopt Model-Agnostic Meta-Learning (MAML) to optimize model initialization for rapid adaptation to unseen tasks under limited data conditions. Extensive experiments are conducted on two benchmark datasets, EMBER and Malicia, under a 5-way 5-shot protocol with disjoint class splits to ensure evaluation on previously unseen malware families. The proposed framework achieves superior performance, reaching 97.8% accuracy on EMBER and 96.2% on Malicia, consistently outperforming state-of-the-art methods. These results indicate that jointly enhancing representation quality and model adaptability can improve classification accuracy and unseen-family performance under the evaluated 5-way 5-shot protocol. Overall, MetaWave provides an effective framework for few-shot malware classification and offers a promising basis for detecting emerging malware under limited-data conditions, while robustness to adversarial perturbation, obfuscation, and polymorphism remains to be validated through dedicated future evaluation. Full article

Few-shot image classification requires models to generalize from limited labeled examples. While metric-based approaches such as Prototypical Networks have demonstrated strong performance, they rely exclusively on visual features and ignore the rich semantic information encoded in class names. This paper presents a systematic empirical study investigating the interaction between visual and semantic modalities in few-shot learning. We present Asymmetric Cross-Modal Prototypical Networks(ACM-ProtoNet), a controlled experimental framework which augments standard prototypical learning with frozen CLIP text encoders to incorporate zero-cost linguistic priors. Our method explicitly models the symmetric relationshipbetween visual and semantic modalities through learnable projection heads that map both image and text features into a shared embedding space. Image and text prototypes are fused via a learnable scalar gate $\alpha \in (0,1)$, allowing adaptive balancing of modalities. Under our experimental setup (frozen CLIP encoders, scalar fusion gate, simple template-based prompts), we observe an asymmetric pattern in comprehensive ablation studies on miniImageNet: cross-modal integration yields a statistically significant improvement in five-shot ($+2.12$ pp, $p=0.03125$, Wilcoxon signed-rank test over five seeds) but not in one-shot ($-0.09$ pp, n.s.) learning. Our key contribution is not achieving state-of-the-art accuracy but rather providing controlled empirical evidence about cross-modal interaction patterns under specific design constraints. Further analysis shows that: (1) structured semantic information is essential—random text features harm performance by $7.4$–$8.1$ percentage points; (2) projection heads provide asymmetric benefits, more critical in one-shot ($-2.85$ pp when removed) than in five-shot learning ($-0.74$ pp); (3) text-only prototypes achieve near-random performance (≈20%), suggesting that semantics alone are insufficient in our setup; (4) shuffled-class-name ablation confirms genuine semantic binding, where randomly permuting class-name assignments causes consistent degradation (five-shot: $-5.74$ pp, $p<0.001$; one-shot: $-3.83$ pp, $p<0.001$ across five seeds). These findings, specific to our simple fusion design, reveal an asymmetric pattern that is equally consistent with two hypotheses: (i) semantic priors may require sufficient visual context to be useful, or (ii) our scalar fusion gate may lack the capacity to leverage text in the extreme low-data regime of one-shot learning. This ambiguity motivates future work with more expressive fusion mechanisms and stronger text representations. Full article

Inverse synthetic aperture radar (ISAR) image classification plays a crucial role in remote sensing, traffic monitoring, and maritime surveillance. However, existing methods often suffer from limited labeled data, degraded image quality, and the insufficient adaptability of conventional loss functions. To address these issues, this paper proposes AdaProtoNet, a few-shot ISAR image classification framework based on a ResNet10 backbone and a combined adaptive and cross-entropy loss function. The model adopts a Prototypical Network architecture that balances feature extraction and class discrimination. A customized multicategory ISAR dataset is constructed through 3D target modeling and simulated radar imaging to support few-shot learning. Within the meta-learning paradigm, AdaProtoNet generates class prototypes by averaging support features and performs classification via Euclidean distance measurement. Experimental results demonstrate that AdaProtoNet achieves higher overall accuracy (OA) and stronger generalization than conventional ISAR classification methods. These findings highlight the effectiveness of adaptive-margin optimization in few-shot learning and provide guidance for the development of next-generation remote sensing recognition systems. Full article

Complex farming environments, breed variation, and the high cost of manual annotation remain major obstacles to robust pig detection, while cross-breed detection under few-shot conditions has been insufficiently explored in previous studies. To address this gap, we propose a few-shot pig detection framework that combines an improved YOLOv7 detector with CycleGAN-based pseudo-sample generation. The detector was enhanced through anchor optimization, Efficient Channel Attention (ECA), and Log-Sum-Exp (LSE) pooling to improve localization and feature discrimination in dense pigsty scenes. In addition, an optimized CycleGAN with perceptual loss was used to generate synthetic Duroc-like pig images to enrich the limited target-domain training set. The framework was evaluated using a two-dataset design: a White Pig Base Dataset was used to establish the source-domain detector and validate the architectural improvements, whereas a Duroc Pig Few-Shot Dataset was used to assess cross-breed adaptation under a 10-shot setting. The experimental results show that the proposed method achieved 98.16% mAP on the White pig dataset and 85.52% mAP on the Duroc Few-Shot Dataset. On the Duroc Few-Shot Dataset, the final framework outperformed Faster R-CNN, CenterNet, and YOLOv8, and also surpassed DCGAN- and SRGAN-based augmentation strategies. These results indicate that the proposed method provides an effective and practical solution for cross-breed few-shot pig detection, with potential value for intelligent livestock monitoring under annotation-limited conditions. Full article

In few-shot action recognition (FSAR), limited annotated data and large scene variations make it difficult for models to learn stable spatial semantics and reliable temporal dynamics. As a result, spatiotemporal representations tend to be weak, and models often fail to focus on discriminative motion regions or capture frame-to-frame changes accurately. Furthermore, the insufficient fusion of local details and global context renders the learned features more susceptible to background noise and scene bias. These issues become more pronounced when background clutter is severe or when different action classes share locally similar segments, leading to unreliable support–query matching and shifted similarity distributions, which ultimately result in class confusion. To address these challenges, we propose a bidirectional adaptive spatiotemporal modeling method integrated with contrastive learning for FSAR. The method constructs attention-guided bidirectional differencing features to model inter-frame variations with semantic alignment, while suppressing background noise. It introduces a local–global interactive channel attention module to strengthen both local and global dynamic representations, and integrates dynamic distance adjustment with hard negative mining during tuple-level matching. This combination imposes contrastive constraints that enhance intra-class compactness and inter-class separability, thereby mitigating interference from cross-class similar segments. Experiments under the standard 5-way 1-shot/5-shot protocol demonstrate consistent improvements across multiple datasets, and the proposed method achieves the best performance under the 5-shot setting while remaining competitive under the 1-shot setting. Full article

The rapid proliferation of low-altitude drones has led to increasingly congested and heterogeneous electromagnetic environments, posing significant challenges to fine-grained spectrum awareness and reliable drone management. Specific emitter identification (SEI), which exploits inherent hardware-dependent radio frequency fingerprints, provides an effective physical-layer solution for emitter-level discrimination. However, practical SEI systems often suffer from two critical issues: extremely limited labeled samples for newly emerging emitters and heterogeneous data distributions collected by geographically distributed receivers with mismatched label spaces. To address these challenges, this paper proposes a heterogeneous federated learning (HFL)-based framework for few-shot specific emitter identification (FS-SEI). The proposed framework decouples feature embedding learning from task-specific classification and enables collaborative representation learning across distributed receivers without sharing raw signal data. A metric learning-based training strategy is adopted, where only the feature embedding models are aggregated in the federated process, effectively alleviating the impact of label space mismatch by utilizing center loss and an improved triplet loss. Moreover, two federated optimization schemes, namely gradient averaging (GA) and model averaging (MA), are systematically investigated to analyze their effectiveness under fully heterogeneous settings. Extensive experiments conducted on a real-world dataset demonstrate that the proposed HFL framework significantly outperforms isolated local training. In particular, the GA-based scheme achieves a few-shot identification performance that closely approaches centralized learning while preserving data privacy and robustness against data heterogeneity. The results validate the effectiveness of the proposed approach for practical FS-SEI in low-altitude drone management scenarios. Full article

Cross-domain few-shot hyperspectral image (HSI) classification is challenging due to limited labeled samples and distribution shifts across sensors and acquisition scenes, which often degrade feature representation and classification performance. This study proposes a lightweight hierarchical CNN–Transformer framework, termed LGDAF-Net (Lightweight Global and Local Dual Attention Fusion Network), for effective cross-domain few-shot HSI classification. The framework progressively enhances spectral–spatial representation through three stages: spectral–spatial feature recalibration, local spatial structure perception, and global contextual modeling. Specifically, a spectral–spatial dual-attention enhancement module (SESA) is introduced to emphasize informative spectral responses and suppress redundancy. A Local Attention Spatial Perception Module (LASPM) is designed to capture fine-grained spatial structures, while a lightweight Transformer-based Global Attention Context Modeling Module (GACM) models long-range spatial dependencies. In addition, kernel triplet loss and domain adversarial learning are incorporated to improve feature discrimination and promote cross-domain feature alignment. Experimental results on three benchmark datasets demonstrate that the proposed method achieves competitive performance compared with existing methods. Full article

Few-Shot Class-Incremental Learning (FSCIL) aims to sequentially learn new classes from very few labelled samples while preventing the forgetting of previously acquired knowledge, which has important practical value for remote sensing scene classification (RSSC). Recent studies have shown that applying a Vision Transformer (ViT) pre-trained on natural image datasets to FSCIL tasks can achieve significantly superior performance. Nevertheless, a substantial domain distribution gap exists between natural images and remote sensing images, which leads to severe performance degradation when such models are directly transferred to RSSC. To address the domain gap alongside FSCIL’s inherent stability–plasticity dilemma and overfitting under data scarcity, we propose a Dynamic Expansion Mixture-of-Experts with Pre-trained Vision Transformer (DEM-ViT) framework. Specifically, to alleviate the domain discrepancy, DEM-ViT incorporates an Adapter-Based Mixture-of-Experts (ABMoE) module, which captures the diverse visual patterns of remote sensing scenes through feature reconstruction in the representation space and collaborative learning among multiple experts. Furthermore, to address the stability–plasticity dilemma in FSCIL, we propose a Dynamic Expert Expansion (DEE) strategy, which progressively expands the model capacity along the incremental sessions. DEE provides sufficient space for learning new knowledge while mitigating the forgetting of previous knowledge. In addition, we propose a Semantic-Guided Feature Alignment (SGFA) method to reduce the risk of overfitting under data-scarce conditions. SGFA leverages textual information to construct robust text prototypes and uses them to calibrate the visual feature space. Extensive experiments across three benchmarks indicate that our framework exhibits highly competitive performance compared with state-of-the-art methods. Full article

Extreme class imbalance is a persistent obstacle for machine learning-driven intrusion detection, as rare but high-impact cyberattacks occur far less frequently than benign traffic in training data. In many real-world cybersecurity datasets, this imbalance becomes extreme, with certain attack types containing a handful of samples, effectively placing the problem in a few-shot learning regime. This paper presents a controlled benchmarking study of Generative Adversarial Network (GAN) objectives for synthesizing minority-class cyberattack data. Using the UWF-ZeekData22 network traffic dataset, each MITRE ATT&CK tactic is framed as a separate binary detection task, and tactic-specific GANs are trained solely on minority samples to generate synthetic attack records. Four widely used GAN variants—Vanilla GAN, Conditional GAN (cGAN), Wasserstein GAN (WGAN), and Wasserstein GAN with Gradient Penalty (WGAN-GP)—are compared under unified training steps and fixed augmentation conditions. The utility of generated data is assessed by evaluating downstream detection performance using five traditional classifiers: Logistic Regression, Support Vector Machine, k-Nearest Neighbors, Decision Tree, and Random Forest. The results indicate that GAN augmentation generally strengthens minority-class detection across tactics and models, reducing false negatives and improving recall consistency, while not systematically harming majority-class performance. However, the effectiveness of each GAN objective varies significantly with data sparsity. Specifically, simpler adversarial objectives often outperform more complex architectures by preserving discriminative feature structure, while heavily regularized models may overly smooth minority-class distributions and reduce separability. Wasserstein-based objectives provide improved training stability, but additional regularization does not consistently translate to better detection performance. Overall, the results demonstrate that in extreme-imbalance settings, GAN effectiveness is governed more by data sparsity and structure preservation than by architectural complexity. These findings establish class-specific generative augmentation as a practical strategy for intrusion detection and provide empirical guidance for selecting appropriate GAN objectives for tabular cybersecurity data under highly imbalanced conditions. Full article

Transportation agencies receive large volumes of free-form public comments describing infrastructure conditions, safety concerns, and service issues. These comments are often processed manually for downstream operational actions, which is time-consuming, inconsistent across reviewers, and difficult to scale, thereby limiting their value for operational decision-making. This study presents a machine learning and Large Language Model (LLM) framework for automated triage of free-form public comments, assigning each report to a three-level hierarchical taxonomy consisting of Category, Subcategory, and Final Decision. The proposed framework uses agency historical data together with retrieval-based evidence, where semantically similar past comments are provided to the LLM as contextual support to better align predictions with agency-specific labeling practices. The framework was evaluated using TF-IDF with Logistic Regression, TF-IDF with Linear SVM, embedding-based kNN with cosine similarity, few-shot LLM prompting, and retrieval-based LLM prompting. Results show that retrieval-based prompting achieved the best overall performance, with the highest accuracy at both the Category and Subcategory levels. At the Final Decision level, retrieval-based prompting slightly outperformed kNN, while few-shot prompting performed worse. Error analysis showed that many misclassifications were semantically plausible alternatives, reflecting the overlap across infrastructure-related complaint categories. When a second candidate label was allowed, further improving performance. Latency analysis also indicated that the framework can process more than 2000 comments in under 30 min, supporting faster and more consistent agency workflows. Full article

Cross-topic stance detection degrades when encoders entangle stance signals with topic-specific vocabulary, causing representations that fail to transfer to unseen targets. Existing methods commit to either topic-invariant or topic-aware representations and apply the same strategy uniformly to every input, sacrificing complementary information. We propose ReDyGait, a three-stage framework that disentangles these two types of signals through dedicated contrastive pre-training and recombines them adaptively at inference time. Stage 1 trains a topic-invariant encoder with supervised contrastive loss over cross-topic positives. Stage 2 trains a topic-contextual encoder with bidirectional pair contrastive loss over within-topic positives; both stages employ topic-aware hard negative mining to prevent shortcut learning. Stage 3 freezes the two contrastive encoders and learns a gating network that produces per-instance weights over invariant, contextual, and base-encoder pathways. On VAST, ReDyGait achieves a macro-averaged F1 of 0.782 in the zero-shot setting and 0.752 in the few-shot setting, improving over the strongest baseline by 1.1 points in both; on SEM16t6 in a leave-one-target-out setup, ReDyGait reaches an average F1 of 0.612. Analysis of the learned gate weights shows that the model shifts toward the invariant pathway for unfamiliar topics and toward the contextual pathway when topic-specific patterns are available, confirming that the disentanglement operates as intended. Full article

Autonomous vehicle navigation requires low-latency and energy-efficient machine learning models capable of operating in dynamic and resource-constrained environments. Conventional deep learning approaches are often unsuitable for real-time deployment on embedded edge devices due to their high computational and memory demands. In this work, we propose a unified TinyML-optimized navigation framework that integrates a lightweight convolutional feature extractor (MobileNetV2) with a metric-based few-shot learning classifier to enable rapid adaptation to unseen driving scenarios with minimal data. The proposed framework jointly combines feature extraction, few-shot generalization, and edge-aware optimization into a single end-to-end pipeline designed specifically for real-time autonomous decision-making. Furthermore, post-training quantization and structured pruning are employed to significantly reduce the memory footprint and inference latency while preserving the classification performance. Experimental results demonstrate that the proposed model achieved a 93.4% accuracy on previously unseen road conditions, with an average inference latency of 68 ms and a memory usage of 18 MB, outperforming traditional CNN and LSTM models in efficiency while maintaining a competitive predictive performance. These results highlight the effectiveness of the proposed approach in enabling scalable, real-time navigation on low-power edge devices. Full article

The application of differential privacy in deep learning often leads to significant performance degradation on class-imbalanced medical datasets. Methods such as adding noise to gradients for differential privacy are effective on large datasets, like MNIST and CIFAR-100, but perform poorly on small, imbalanced medical datasets, like HAM10000 and ISIC2019. This is because the imbalanced distribution causes the gradients from the few-shot classes to be clipped, resulting in the loss of crucial information, while the majority classes dominate the learning process. This leads the model to fall into suboptimal solutions early. To address this issue, we propose SDD-DPSGD, which uses a step-wise dynamic exponential scheduling mechanism for noise and clipping thresholds to preserve gradient information. By allocating more privacy budget and employing higher clipping thresholds during the initial training phases, the model can avoid suboptimal solutions and improve its performance. Experiments show that SDD-DPSGD outperforms comparable algorithms on the HAM10000 dataset, and the ISIC2019 dataset. Full article