Sparse grid quadrature in high dimensions with applications in finance and insurance (Q608138): Difference between revisions

This monograph mainly presents results of the author and his coauthors on developing methods of numerical computation of high-dimensional integrals (of hundreds of variables) \[ I_{\varphi}g:=\int_{R^d}g(z)\varphi(z)dz \] over the \(d\)-dimensional Euclidian space with the Gaussian weight function \(\varphi\) which can be transformed into integrals over the unit cube \[ If:=\int_{[0,\;1]^d}f(x)dx. \] Such high-dimensional integrals arise under the discretization of finance and insurance dynamic stochastic models representing risk or revenue. The main methods presented are based on the analysis of variance (ANOVA) of the integrand. There the function \(f\) is decomposed into a sum of \(2^d\) terms describing the relative importance of a subset of variables with respect to the total variance of \(f\). Together with a classical ANOVA using the Lebesgue measure \(d\mu(x)=dx\) the anchored-ANOVA with the Dirac measure \(d\mu(x)=\delta(x-a)dx\), \(a\in[0,1]\) is used. The notions of effective dimension (reducing calculations in the common case) are introduced in both cases. Corresponding error bounds for approximation and integration are derived. The proposed dimension-wise quadrature methods include sparse grid methods as a special case. They are implemented for following next applications from finance and insurance: interest rate derivatives, path-dependent options, performance-dependent options, and asset-liability management in life insurance.