750 papers
This study demonstrates that a lightweight YOLO11n deep learning model can achieve near-perfect accuracy in automatically classifying dried sea cucumber processing methods (salting, smoking, and roasting) from dense-object images, offering a robust, non-destructive alternative to subjective manual inspection for quality grading.
This paper proposes SGMS-DTDNet, a semantic-guided multi-scale dual-teacher distillation network that integrates domain-specific semantics, structural priors, and adaptive feature selection to achieve state-of-the-art performance in unified medical and industrial anomaly detection by overcoming challenges like scarce abnormal samples and heterogeneous defect morphologies.
This study demonstrates that parameter-efficient fine-tuning (PEFT) strategies, such as partial fine-tuning and BitFit, outperform full fine-tuning in electrocardiogram foundation models by significantly enhancing generalizability and reducing overfitting, particularly on out-of-domain data and resource-constrained scenarios.
This paper introduces a constraint-guided framework that repairs deployed neural networks by localizing fault-relevant weights via DeepSHAP, collecting symbolic constraints through concolic testing, and optimizing updates with Max-SMT, demonstrating through extensive experiments that systematic design-space exploration reveals critical repair strategies—such as bias-only modifications—that significantly reduce adversarial and backdoor vulnerabilities while preserving model fidelity.
This systematic review of 32 studies (2022–2025) identifies four hybrid Reinforcement Learning and Metaheuristic architectures for independent UAV path planning, finding that while these frameworks improve adaptability and convergence, their real-world readiness is currently limited by a complete lack of physical hardware validation, code reproducibility, and rigorous statistical testing.
This paper proposes an integrated AIOps framework that combines XGBoost-based predictive risk scoring, autoencoder-driven anomaly detection, and Random Forest-based automated remediation to proactively mitigate deployment failures and significantly reduce mean time to recovery in CI/CD environments.
This paper presents a progressive multi-architecture computer vision framework for real-time defect detection in Wire Arc Additive Manufacturing (WAAM), systematically evaluating and comparing interpretable two-step models, end-to-end deep learning networks, and a novel learnable preprocessing architecture (MCFE-Net) to optimize the trade-offs between computational efficiency, interpretability, and anomaly detection accuracy.
This empirical study tests a consent-friction functional in multi-agent reinforcement learning, rejecting its specific monotone form as a universal predictor while demonstrating that observed coordination effects are largely artifacts of design choices and that the sole surviving structural effect—where cooperative alignment improves outcomes—is driven by increased feasibility in shared-resource environments rather than reduced learning friction.
This paper presents a comprehensive comparative study of machine learning and deep learning approaches for multilingual speech accent recognition using the Speech Accent Archive and AccentDB datasets, demonstrating that hybrid CNN–BiLSTM models achieve superior accuracy (up to 99.18%) and highlighting the feasibility of deploying these systems for personalized, accent-aware interactions in consumer electronics.
This paper proposes an adaptive E-tutoring system that integrates Atanassov's Intuitionistic Fuzzy Logic with an ontology-based domain model to effectively handle uncertainty in student knowledge assessment, enabling personalized teaching materials and dynamic feedback through a functional Python prototype.