RESEARCH · GLOBAL
Randomized neural networks accelerate CVA estimation for American equity options
Study demonstrates randomized neural networks (fast, no gradient descent) outperform traditional Monte Carlo for computing Credit Valuation Adjustment (CVA)—the cost of counterparty default—on American options under both Black-Scholes and Heston dynamics.
WHY IT MATTERS
CVA is a regulatory capital requirement; faster, accurate neural-net estimation reduces compute bottlenecks in risk engines and improves intraday trading desk P&L reporting.
Source: arXiv · 2026-06-24