Amir Malekesfandiari | Searcharxiv

arXiv2026

Logical Grammar Induction via Graph Kolmogorov Complexity: A Neuro-Symbolic Framework for Self-Healing Clinical Data Integrity

The reliability of Healthcare Information Systems (HIS) is frequently compromised by human-induced data entry errors, which existing statistical anomaly detection methods fail to distinguish from legitimate clinical extremes. This paper proposes Logic-GNN, a novel neuro-symbolic framework that treats clinical records as a structured ``private language'' governed by latent logical games. By integrating Temporal Graph Neural Networks (TGNN) with Graph Kolmogorov Complexity, we induce a symbolic grammar that represents the underlying logic of medical interactions. We define anomalies as ``grammatical violations'' that cause a significant expansion in the Minimum Description Length (MDL) of the clinical graph. Evaluated on the Sina System dataset (2M+ records), Logic-GNN achieves an F1-score of 0.94, outperforming state-of-the-art baselines by 12\% in distinguishing between life-threatening medical outliers and data corruption. Our approach introduces a self-healing mechanism that suggests logical corrections to maintain data integrity in real-time HIS environments.

cs.LG↗

arXiv2025

Multimodal Framework for Explainable Autonomous Driving: Integrating Video, Sensor, and Textual Data for Enhanced Decision-Making and Transparency

Autonomous vehicles (AVs) are poised to redefine transportation by enhancing road safety, minimizing human error, and optimizing traffic efficiency. The success of AVs depends on their ability to interpret complex, dynamic environments through diverse data sources, including video streams, sensor measurements, and contextual textual information. However, seamlessly integrating these multimodal inputs and ensuring transparency in AI-driven decisions remain formidable challenges. This study introduces a novel multimodal framework that synergistically combines video, sensor, and textual data to predict driving actions while generating human-readable explanations, fostering trust and regulatory compliance. By leveraging VideoMAE for spatiotemporal video analysis, a custom sensor fusion module for real-time data processing, and BERT for textual comprehension, our approach achieves robust decision-making and interpretable outputs. Evaluated on the BDD-X (21113 samples) and nuScenes (1000 scenes) datasets, our model reduces training loss from 5.7231 to 0.0187 over five epochs, attaining an action prediction accuracy of 92.5% and a BLEU-4 score of 0.75 for explanation quality, outperforming state-of-the-art methods. Ablation studies confirm the critical role of each modality, while qualitative analyses and human evaluations highlight the model's ability to produce contextually rich, user-friendly explanations. These advancements underscore the transformative potential of multimodal integration and explainability in building safe, transparent, and trustworthy AV systems, paving the way for broader societal adoption of autonomous driving technologies.

cs.MM↗