XAI-Based Sensitivity Analysis of High-Speed Link Considering Channel Operating Margin

In this paper, we propose explainable artificial intelligence (XAI)-based high-speed link sensitivity analysis (SA) method considering channel operating margin (COM) to provide a fast, accurate, and interpretable method for analyzing highspeed links. By computing Shapley additive explanations (SHAP) values for each design parameter, our method reveals not just their relative importance but also the direction of their influenceindicating whether adjustments in parameters lead to over- or under-equalization (EQ). We further validate our approach by comparing it to a conventional Morris method, demonstrating the advantages of SHAP in capturing context-dependent effects and detailed parameter interactions. The proposed methodology enables designers to analyze system performance more efficiently, reduce time-consuming manual tuning, and gain valuable insights into complex, nonlinear EQ behaviors in high-speed communication systems.