924 papers
This paper introduces the Structure-Aware Collaborative Generative Framework (SACGF), a novel pipeline integrating Stable Diffusion, ControlNet, and SAM to achieve high-fidelity digital reconstruction of lacquer art surfaces by balancing structural consistency, texture fidelity, and style coherence, validated by a new quantitative metric and a specialized dataset.
This paper proposes SEM-EDGE-X, an intent-aware semantic edge intelligence framework that integrates semantic feature extraction, adaptive routing, and multi-agent coordination to significantly outperform conventional communication models in low-latency, dynamic 6G UAV–vehicular networks.
This paper demonstrates that machine learning models, particularly Recurrent Neural Networks, can accurately predict waiting times in complex, time-varying priority queueing systems using only queue length and customer priority, outperforming tree-based methods across non-preemptive and various preemptive service rules.
This study evaluates six deep learning architectures for maritime object detection across diverse weather conditions, demonstrating that the Vision Transformer (ViT) model outperforms CNN-based alternatives by achieving 100% accuracy, the lowest error rates, and the fastest processing speed, thereby highlighting its potential for enhancing AI-driven maritime security and surveillance.
This paper proposes a novel feature selection method that integrates kernel alignment to unify heterogeneous medical data within a Reproducing Kernel Hilbert Space and extends three-way clustering to explicitly model boundary uncertainty, thereby effectively reducing dimensionality while significantly improving disease prediction accuracy and stability on multimodal clinical datasets.
This paper proposes a weakly supervised framework for transaction-graph anomaly detection that leverages optimal transport to align hierarchical rule semantics, derived from decision trees, with continuous transaction representations, thereby generating high-quality pseudo-labels and outperforming existing baselines on benchmark datasets.
This paper argues that current public regulatory records for AI medical devices are inadequate for lifecycle oversight due to a lack of structured, queryable data on software changes and AI-specific attributes, and proposes a new data specification to address these critical deficits across major global jurisdictions.
This paper proposes a multimodal unsupervised anomaly detection framework that fuses textual semantics from OCR and financial logical features to effectively identify structural irregularities and fraud in semi-structured invoices, outperforming single-modality baselines on real-world data.
This paper introduces CLASP, an adaptive constrained reinforcement learning scheduler that enforces explicit latency constraints via a PID-controlled Lagrangian price and price-conditioned policy, thereby maintaining stable energy efficiency and violation targets across varying workloads where traditional fixed-weight reward approaches fail.
This paper proposes and evaluates a multi-layer adaptive Zero Trust framework that utilizes a context-aware trust scoring mechanism and rigorous mathematical modeling to demonstrate a significant reduction in attack success rates and lateral movement within cloud environments compared to traditional perimeter-based security.