Valuation of financial derivatives with constrained Gaussian process regression

Details

Supervisors

Hainaut, Donatien

Faculty

Faculté des sciences

Degree label

Master [120] en sciences actuarielles, à finalité spécialisée

Abstract

This thesis studies Constrained Gaussian Process Regression (CGPR) for option pricing. The study puts the focus on a practical problem in quantitative finance where prices and Greeks often have to be computed many times for the same model. In practice, standard methods such as FFT or Monte Carlo are commonly used to obtain these prices. However, these approaches do not always provide a reusable pricing surface and greeks that can easily be evaluated again. The approach studied here is different because CGPR uses the Feynman–Kac PDE itself to approximate the pricing function, by imposing the PDE residual at interior points and the payoff or boundary conditions at boundary points.