2718 papers
This paper presents an interpretable multimodal machine learning framework that combines CBC and HPLC biomarkers to achieve highly accurate (99.89%) and clinically consistent multiclass classification of thalassemia using a stacking ensemble model, with SHAP analysis confirming the top predictive features align with standard diagnostic criteria.
This study evaluates a bilingual retrieval-augmented generation system built on ANVISA regulatory documents, demonstrating its high accuracy in locating, citing, and summarizing Portuguese and English regulatory requirements for drug development while effectively handling unanswerable queries after calibration.
This paper proposes a CLCM model, integrating Conv1D-LSTM, CNN, and a multi-head attention mechanism, to effectively assess security risks in virtual reality learning systems within the Unmanned Swarms Internet of Things by accurately analyzing heterogeneous data modalities and overcoming limitations of existing models.
This paper proposes an automated tomato leaf disease diagnosis framework using EfficientNetB3 transfer learning and Grad-CAM explainable analysis, which achieves high classification accuracy and interpretability for detecting Early Blight, Late Blight, Leaf Mold, and healthy leaves to support precision agriculture.
This paper introduces DYNAMICS-WEB, a pre-registered measurement demonstrating that a distinct "character" signal within the web's link graph, which is largely independent of authority and topic, can subconsciously influence reader conversion rates when aligned with user preferences.
This study proposes a hybrid MSCNN-LSTM model enhanced with Squeeze-and-Excitation attention to predict hourly Air Quality Index values, demonstrating superior accuracy and stability compared to traditional deep learning models through multi-scale feature extraction and adaptive channel weighting on data from Yiyang City.
NeviUNet is an efficient, attention-enhanced U-Net variant that achieves accurate and low-latency skin lesion segmentation in total body images by integrating Mish activation, Attention Gates, and SE blocks while omitting computationally expensive components.
The paper presents FlapFlowNet, a Transformer-based streaming framework that achieves high-accuracy, real-time perfusion monitoring for variable-length flap surgery sequences by combining a lightweight CNN, temporal modeling, and a medical-optimized loss function to significantly reduce critical false-negative rates.
The paper presents ORACOLO, a neuro-symbolic question-answering architecture that integrates ontology-constrained SPARQL query generation with dense retrieval fallback, demonstrating significantly improved factual accuracy over pure dense retrieval while maintaining comparable faithfulness.
This study proposes a hybrid CNN-LSTM deep learning model trained on a custom dataset combining global benchmarks and NS-3 simulated SDN traffic to achieve 97.8% accuracy in detecting cyberattacks against Sudan's e-government infrastructure, offering a scalable solution to enhance national cybersecurity.