2608 papers
This paper proposes an adaptive cascaded fast partitioning algorithm for Versatile Video Coding (VVC) that integrates visual perceptual analysis and multi-level machine learning to significantly reduce encoding complexity while maintaining high coding efficiency.
This study introduces an Explainable Dual-view Workflow Discovery framework that analyzes log data from 278 students to reveal that effective GenAI-supported learning is characterized not merely by chat frequency, but by the integration of AI feedback into writing, revision, and cross-resource regulation.
This paper proposes SQCAFusion, an illumination-adaptive infrared and visible image fusion framework that employs multi-scale scene-query cross-attention with learnable scene tokens to dynamically modulate features during early encoding, thereby resolving scene-blindness issues and enhancing fusion quality under varying lighting conditions.
This paper introduces a cache-line cost model for open addressing without reordering, demonstrating that while asymmetric bucketing achieves optimal memory access bounds of , symmetric approaches are significantly worse and probe-optimal hierarchical schemes remain cache-suboptimal due to unavoidable memory-access costs dictated by the parameter .
WeanNet proposes a parameter-efficient transfer learning method that uses a decaying scalar gain to temporarily connect a frozen parent network to a new student, enabling lossless parent removal and reducing model size while demonstrating competitive performance on concept-drift tasks compared to Progressive Neural Networks, though its general superiority remains limited by specific task and configuration dependencies.
This paper introduces ResilientNet, a reliability-aware framework that enhances Deep Neural Network resilience against hardware faults through four synergistic design and training strategies—bounded activations, layer reordering, NaN filtering, and curriculum-based fault training—demonstrated to significantly reduce error propagation and maintain high classification accuracy with negligible inference overhead across extensive fault injection experiments.
This paper proposes a MoE-CB framework, integrating a gating mechanism with hybrid CNN-LSTM/CNN-BiGRU experts and a CatBoost regressor, to enable adaptive and intelligent dynamic resource allocation in 5G networks with high prediction accuracy.
This paper presents a new, pairing-free partially blind strong designated-verifier signature scheme that ensures fine-grained privacy and non-transferability with low computational overhead, making it highly suitable for resource-constrained edge environments like IoT and vehicular networks.
This study proposes an enhanced memetic algorithm within a fuzzy-CPT framework to optimize emergency outpatient physician scheduling by simultaneously minimizing patient psychological distress and physician fatigue perception, validated through real-world data from a Shanghai hospital during the COVID-19 pandemic.
This paper introduces MSCE, a multi-signal post-hoc calibration framework that significantly improves the reliability of field-level confidence estimates in business document extraction by fusing diverse signals like OCR quality and model uncertainty, thereby enabling higher automation rates with strict precision requirements.