2787 papers
This paper introduces a novel, low-code Shiny-based data anonymization platform that empowers researchers, particularly in resource-constrained settings, to protect sensitive information and ensure data utility through an intuitive interface, integrated risk assessment, and automatic generation of reproducible code in R, Stata, and Python.
This paper introduces a novel recurrent neural network inspired by exponential smoothing methods, which utilizes a particle swarm optimization-based learning algorithm and a new objective function to achieve superior forecasting performance for seasonal time series compared to existing deep and shallow artificial neural networks.
This paper introduces a unified, multi-dimensional survival benchmark for temporal dropout risk prediction using the OULAD dataset, demonstrating that behavioral and temporal signals outperform static attributes while highlighting the necessity of separating model families to avoid evaluation artifacts.
This paper proposes a robust multimodal deep learning framework that fuses facial expressions, heart rate variability, and speech signals using CNNs, BiLSTMs, and ECAPA-TDNNs with a self-organizing map for feature fusion, achieving 98.6% accuracy in recognizing seven emotional states and significantly outperforming unimodal baselines.
This study proposes a multimodal framework that leverages human dance motion dynamics, extracted from pose landmarks in dance videos, to enhance personalized music recommendation by better aligning suggested tracks with users' embodied rhythmic preferences and affective states.
This study reveals that visual distinctiveness, rather than rarity alone, is the primary driver of memorization and membership inference vulnerability in fine-tuned medical imaging models, demonstrating that medical-domain pretraining fails to mitigate these risks while DP-LoRA effectively reduces them.
This paper proposes Feature Calibration for Camera Bias Elimination (FCBE), a framework that mitigates camera-induced bias in unsupervised person re-identification through style-aware metric calibration and distribution-aware median calibration, thereby enhancing clustering reliability and achieving state-of-the-art performance on Market-1501 and MSMT17 datasets.
This paper proposes a deterministic, decentralized Double-Sided Walrasian Auction mechanism for collaborative Multi-Access Edge Computing that replaces computationally expensive and non-adaptive AI-based approaches with exact KKT-optimal bidding and a discrete-time price-adjustment procedure, achieving superior execution speed, energy bounds, and social welfare maximization even under high network loads.
This measurement note audits the discrepancy between nominal compression ratios and the actual reader-visible context retained in a capped FLAN-T5 QA pipeline, demonstrating that while higher nominal ratios increase pre-wrap token counts, the final model input is significantly constrained by template overhead and truncation, and that content selection quality matters more than token budget alone.
This paper presents a physics-constrained neural identification framework that simultaneously reconstructs thermal conversion curves and accurately estimates Arrhenius kinetic parameters from multi-rate thermogravimetric data, effectively overcoming the ill-conditioning and noise sensitivity of conventional methods to enable robust prediction across unseen heating rates.