rho_grid= -1:0.1:1;
Nrho = length(rho_grid);
for i = Nrho : -1 : 1
rho = rho_grid(i);
Rq = @(rho) chol(value(Q(rho)),'upper');
Rp = @(rho) chol(value(P(rho)),'lower');
pro = @(rho) Rq(rho)*Rp(rho);
[Uvals(:,:,i), Svals(:,:,i), Vvals(:,:,i)] = svd(pro(rho));
end
U = @(rho) Uvals(:,:,interp1(rho_grid, 1:Nrho, rho, 'nearest'));
S = @(rho) Svals(:,:,interp1(rho_grid, 1:Nrho, rho, 'nearest'));
V = @(rho) Vvals(:,:,interp1(rho_grid, 1:Nrho, rho, 'nearest'));
Instead of using 'nearest' in computing the index, you could go for something more sophisticated such as determining the mixing between the two closest values, and doing a weighted calculation to interpolate at the rho.
Best Answer