Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.798516
Title: Joint distribution of passage times of an Ornstein-Uhlenbeck diffusion and real-time computational methods for financial sensitivities
Author: Jiang, Yupeng
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
This thesis analyses two broad problems: the computation of financial sensitivities, which is a computationally expensive exercise, and the evaluation of barriercrossing probabilities which cannot be approximated to reach a certain precision in certain circumstances. In the former case, we consider the computation of the parameter sensitivities of large portfolios and also valuation adjustments. The traditional approach to compute sensitivities is by the finite-difference approximation method, which requires an iterated implementation of the original valuation function. This leads to substantial computational costs, no matter whether the valuation was implemented via numerical partial differential equation methods or Monte Carlo simulations. However, we show that the adjoint algorithmic differentiation algorithm can be utilised to calculate these price sensitivities reliably and orders of magnitude faster compared to standard finite-difference approaches. In the latter case, we consider barrier-crossing problems of Ornstein-Uhlenbeck diffusions. Especially in the case where the barrier is difficult to reach, the problem turns into a rare event occurrence approximation problem. We prove that it cannot be estimated accurately and robustly with direct Monte Carlo methods because of the irremovable bias and Monte Carlo error. Instead, we adopt a partial differential equation method alongside the eigenfunction expansion, from which we are able to calculate the distribution and the survival functions for the maxima of a homogeneous Ornstein-Uhlenbeck process in a single interval. By the conditional independence property of Markov processes, the results can be further extended to inhomogeneous cases and multiple period barrier-crossing problems, both of which can be efficiently implemented by quadrature and Monte Carlo integration methods.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.798516  DOI: Not available
Share: