Global Sensitivity Analysis

Global sensitivity analysis (GSA) quantifies how variations in model inputs affect outputs, aiming to identify the most influential factors and their interactions. Current research emphasizes developing more efficient and robust GSA methods, particularly for high-dimensional and mixed-variable (both categorical and numerical) data, often employing techniques like Sobol indices, Shapley values, and Gaussian processes within active learning frameworks. These advancements are crucial for improving model explainability, optimizing experimental design, and mitigating bias in various fields, including machine learning, engineering design, and risk assessment.