# Combine griddedInterpolants on the same grid

5 views (last 30 days)
Konrad Warner on 20 Oct 2020
Edited: Konrad Warner on 20 Oct 2020
I'm wondering if there is a way to combine griddedInterpolants on the same grid.
I have uniform ngrids (201-by-201) for x and y and three different z-grids (surfaces) 201-by-201-by-3.
To querry all three z-values for a x-y-pair I can loop through three Interpolants but is there an faster/ smarter way?
% Show plots
surf(xGrid,yGrid,zGrid(:,:,1))
surf(xGrid,yGrid,zGrid(:,:,2))
surf(xGrid,yGrid,zGrid(:,:,3))
% Interpolant for each z-surface
C(1).F = griddedInterpolant(xGrid,yGrid,zGrid(:,:,1));
C(2).F = griddedInterpolant(xGrid,yGrid,zGrid(:,:,2));
C(3).F = griddedInterpolant(xGrid,yGrid,zGrid(:,:,3));
% Querry points, eg
x = 4;
y = -1;
for i = 1 : 3
out(i) = C(i).F(x,y);
end

Steve Eddins on 20 Oct 2020
I don't think so. If you have a large number of query points, you might try replacing the Values property of the griddedInterpolant for each of the three planes in your grid, instead of creating three different griddedInterpolant objects.
F = griddedInterpolant(xGrid,yGrid,zGrid(:,:,1));
out{1} = F(xx,yy);
F.Values = zGrid(:,:,2);
out{2} = F(xx,yy);
F.Values = zGrid(:,:,3);
out{3} = F(xx,yy);
I'm not sure if this will actually execute faster, though.
The MATLAB Math team is aware of this workflow and is considering ways to improve it.

#### 1 Comment

Konrad Warner on 20 Oct 2020
Ok thanks anyway

Bruno Luong on 20 Oct 2020
If you know in advande the reference gridded points and the query points, (but the z data change), you migh build one ce interpolation matrix, then multiply by the data to retrieve the interpolation values. See this thread

#### 1 Comment

Konrad Warner on 20 Oct 2020
It's more that the zData are constant and the querry points change.
It's a classifier: Each surface (3rd dimension) represants a class and I want to find the degree of membership (zValues) to each class for the Features x and y.