2608 papers
This study demonstrates that rumors and non-rumors exhibit distinct early propagation structures, characterized by stronger mesoscale topological cavities and divergent cluster-merging behaviors, which can be effectively identified using persistent homology on 3D point clouds to improve prediction accuracy.
This paper proposes a hybrid fault diagnosis framework for distribution systems that integrates a dynamic threshold multi-scale convolutional autoencoder with Mahalanobis distance-enhanced K-nearest neighbor and a dynamic energy score threshold strategy to overcome feature overlapping challenges and accurately locate adjacent-node faults under complex noise environments.
This paper introduces DA-GAT-CADS, a learning-based solver for the Euclidean Traveling Salesman Problem that combines a geometry-anchored Delaunay graph encoder with a context-adaptive, gate-controlled dynamic sampling decoder to effectively balance computational efficiency and solution quality by balancing local structural priors with state-dependent nonlocal candidate selection.
This study presents and validates a Type-2 Fuzzy Expert System developed in Python that leverages expert knowledge to effectively diagnose breast cancer and recommend treatments, achieving high accuracy (98%) and robust performance metrics to address diagnostic uncertainty and improve patient care in resource-limited settings.
This paper proposes a novel hybrid framework combining a Temporal Convolutional Network (TCN) with an improved Transformer featuring segmented sparse attention and MAML-based meta-learning to achieve high-accuracy, computationally efficient, and robust small-sample remaining useful life prediction for rolling bearings under complex operating conditions.
This paper proposes Ultra-Fusion, a multimodal dynamic detection framework that integrates a novel RGB fingerprint mapping algorithm, a hybrid CNN-Mamba-Transformer engine, and a statistical branch to achieve state-of-the-art accuracy and robust generalization in classifying polymorphic and metamorphic malware.
The paper introduces CaliCrack, a parameter-free training-time loss function that penalizes boundary inflation in crack segmentation models, significantly improving precision and F1 scores on the DeepCrack dataset by reducing over-thick predictions that lead to false positives in automated road inspection.
This study benchmarks six large language models against human experts for AGREE II appraisal of sepsis clinical practice guidelines, revealing that while models achieve moderate agreement, they exhibit consistent domain-specific biases and varying efficiency, suggesting their optimal role is as triage tools within a human–AI hybrid workflow rather than as standalone evaluators.
This paper benchmarks multiple machine learning models for 5G NR PRACH interference detection, revealing that while Transformer Encoders offer high precision, CNN and XGBoost provide the optimal balance of high accuracy and computational efficiency, making them the most suitable candidates for real-time interference management.
This Design-Based Research study details the development and field testing of an Excel-VBA validation tool for the Mozilla Common Voice platform, demonstrating how iterative design improvements that prioritize community orthography preferences over technical features can enhance data quality and participant sustainability in endangered Circassian language projects.