This paper introduces ECGPD-LEF, a novel framework that integrates foundation model-derived diagnostic probabilities with interpretable modeling to detect low left ventricular ejection fraction (LEF) from ECGs. It achieves robust performance, outperforming existing black-box models, while providing clinical interpretability through identifiable high-impact predictors.

This paper introduces FL-PBM, a novel client-side pre-training defense mechanism for Federated Learning to mitigate backdoor attacks. It proactively identifies and neutralizes poisoned data using PCA, GMM clustering, and adaptive blurring before local model training, significantly reducing attack success rates while maintaining high model accuracy.

Trustworthy AI systems increasingly rely on probabilistic computation, shifting performance bottlenecks from arithmetic to memory, which must deliver both data and randomness. This paper introduces a unified data-access perspective, treating deterministic access as a limiting case of stochastic sampling, to analyze and address these new memory challenges.

This review introduces a framework that combines deep learning with committor analysis and explainable AI (XAI) to systematically identify reaction coordinates in complex molecular systems. The approach enables the quantitative assessment of individual input variable contributions, enhancing the interpretability of molecular transition pathways.

This paper explores lightweight Generative AI (GenAI) models for network traffic synthesis to address data scarcity and privacy in Network Traffic Classification (NTC). It evaluates transformer-based, state-space, and diffusion models, demonstrating their effectiveness in generating high-fidelity synthetic traffic for training and augmenting NTC systems.