D2H-AD: A Hybrid Model Utilizing Hyperdimensional Computing for Advanced Anomaly Detection

Short summary

D2H-AD is a novel anomaly detection framework using Hyperdimensional Computing that combines distance-based similarity and density-aware encoding. It achieves 5.4% higher ROC-AUC than existing methods while offering lightweight, interpretable operation—ideal for edge AI and TinyML deployments.

• •Combines distance-based similarity and density-aware encoding within a unified HDC framework
• •5.4% higher ROC-AUC over existing methods; outperforms five established baselines across all datasets
• •Lightweight with small memory footprint and low-latency binary computations for resource-constrained environments

Generated with AI, which can make mistakes.