Generative AI Enables Bayesian-Calibrated Sensitivity Analysis
Xuyuan Wang publishes a generative modeling framework for global sensitivity analysis of Bayesian-calibrated models in PLOS Computational Biology on March 16, 2026. It employs autoregressive Rosenblatt transforms and diffusion-based Shapley estimators to model high-dimensional posterior correlations and demonstrates scalability and improved data-relevant sensitivity estimates on COVID-19 transmission and cancer immunotherapy case studies.
Key Points
- 1Introduces generative modeling framework for GSA on Bayesian posterior distributions with correlated parameters.
- 2Employs autoregressive Rosenblatt transforms and diffusion-based Shapley estimators to avoid restrictive dependence assumptions.
- 3Enables practitioners to compute data-relevant Sobol and Shapley sensitivities scalable to complex biological models.
Scoring Rationale
Strong methodological novelty with peer-reviewed validation and reusable code, limited by demonstrations confined to two biological models.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
