927 papers
This paper presents a knowledge-constrained neural framework that embeds formal geometric rules into a graph-completion model to guarantee exact rotational symmetry and structural validity in Islamic geometric patterns, demonstrating that while data augmentation improves fidelity, only symmetry-structured inference ensures zero violations.
GeoMind is an autonomous multi-agent framework that integrates a local LLM, RAG-based context injection, and a deterministic geospatial validator within a closed-loop self-correction system to achieve a state-of-the-art 90% task success rate and significantly improved spatial accuracy in geospatial analysis, effectively overcoming the semantic and geographic hallucinations common in general-purpose LLMs.
This paper presents a statistically validated benchmark of Multilayer Perceptron, XGBoost, and Random Forest models for contactless respiratory pattern classification using Software-Defined Radio, demonstrating that MLP and Random Forest achieve superior accuracy over XGBoost while identifying fast and sleep-apnea patterns as the most challenging to distinguish and highlighting Random Forest's higher generalization gap due to overfitting.
This paper presents an AI-powered Smart Physiotherapy Assistant that utilizes computer vision and machine learning to provide real-time posture analysis and corrective feedback via a web interface, thereby enabling accessible, cost-effective, and self-paced rehabilitation without constant therapist supervision.
This study presents LGBM-BBB, a high-speed and accurate web server for predicting blood-brain barrier permeability that utilizes LightGBM trained on an optimized set of 59 features with winsorized outlier handling and Matthew Correlation Coefficient threshold optimization to facilitate efficient neurotherapeutic drug discovery.
This paper introduces a comprehensive framework that combines LSTM-based knot generation, stochastic growth modeling, and conditional diffusion models to create realistic synthetic Norway spruce boards with known internal structures, thereby enabling the training of deep-learning systems for wood assessment tasks where real paired data is scarce.
This paper introduces MobileRACMA, a lightweight and reliability-aware cross-modal attention architecture that fuses RGB and thermal imagery to achieve robust, metric monocular depth estimation on edge platforms while formally guaranteeing bounded information exchange to prevent model instability under photometric degradation.
This study assesses the critical lack of enterprise readiness for quantum computing threats, highlighting that most systems remain vulnerable and migration planning is minimal, and proposes the Quantum-Resilient Security Architecture (QRSA) Framework to guide organizations in transitioning to NIST-standardized post-quantum cryptographic protocols.
This paper presents a comprehensive survey of Physics-Informed Transformers (PITs) based on 54 recent studies, systematically analyzing their architectural innovations, superior spatiotemporal modeling capabilities, and diverse applications while highlighting their data efficiency and outlining future challenges in scalability and benchmarking.
This study employs a machine-readable knowledge-graph audit to conclude that gazetamercantil.com is the sole documented digital legacy and canonical contemporary domain of the 1920-founded Brazilian newspaper Gazeta Mercantil, as it is the only endpoint satisfying six cumulative criteria across public bibliographic, institutional, and scholarly infrastructures.