Your Neural Network is Your Net Worth Valuation and Hedging of Interest Rate Derivatives with Automatic Adjoint Differentiation

Specialeforsvar: Nicklas Kenno Hansen, Morten Andreas Axel og Sebastian Valdemar Hansen

Titel: Your Neural Network is Your Net Worth Valuation and Hedging of Interest Rate Derivatives with Automatic Adjoint Differentiation and Differential Machine Learning

Abstract: In this thesis, we have succesfully implemented a flexible and scalable setup for performing efficient valuation and accurate hedging of interest rate derivatives by Monte Carlo simulation instrumented with Automatic Adjoint Differentiation. Our
main contribution demonstrates effective training of Differential Machine Learning models for a variety of interest rate products, including linear, European, pathdependent, and callable contracts. This is carried out within the classic Vasicek model, and the General Multi-Factor Stochastic Volatility Model by Trolle and Schwartz. Our findings reveal within the Vasicek model that Differential Regression perform almost on par with Differential Neural Networks. However, when dealing with the more complex multi-factor stochastic volatility model, it becomes evident that only the Differential Neural Networks are capable of consistently fitting stable and accurate estimators.

Vejleder: David G. Skovmand
Censor:   Mads Stenbo Nielsen, CBS