Deep Rough Volatility Calibration: Posterior Uncertainty Quantification and Information-Theoretic Explainability
ORIGINAL / Uncertainty and Explainability in Deep Rough Volatility: A Neural Information-Theoretic Posterior Approach
This study proposes a simulation-based inference framework for rough Heston calibration that learns the posterior distribution of parameters conditional on an implied-volatility surface and propagates posterior uncertainty through heteroscedastic neural surrogate pricers to produce uncertainty-aware price intervals; it also introduces Hellinger-SHAP to explain posterior information gain. Its novelty lies in combining posterior inference with explainability to provide a transp
01 ABSTRACT
Fact: The authors develop a simulation-based inference framework for rough Heston calibration, use neural ratio estimation to obtain posterior samples, propagate them through heteroscedastic surrogate pricers to form posterior-predictive distributions and price intervals, and propose Hellinger-SHAP, which applies local-background Kernel SHAP to a posterior-information functional. Simulation studies show that posterior-predictive intervals provide calibrated or conservative coverage for forward-start, barrier, and realized-variance claims, while point plug-in prices can be materially unreliable in selected contract regimes. Author view: The UQ and XAI analyses provide a transparent framework for uncertainty-aware neural calibration and downstream exotic pricing, but conclusions are limited to the specified prior-predictive model.
02 KEY FINDINGS
- Proposes a simulation-based inference framework for rHeston calibration that learns the posterior distribution of parameters conditional on an IV surface.
- Propagates posterior samples through heteroscedastic neural surrogate pricers to obtain posterior-predictive price intervals combining parameter and surrogate uncertainty.
- Introduces Hellinger-SHAP, applying local-background Kernel SHAP to a posterior-information functional to identify maturity–moneyness regions associated with posterior information gain for each parameter.
AI GENERATED SUMMARY / DISCOVERED BY ARXIV Q-FIN