## Is there any more effiecient way to implement this?

### Dimitrios (view profile)

on 18 Oct 2014
Latest activity Edited by per isakson

### per isakson (view profile)

on 18 Oct 2014

Using MATLAB:I have a table of lets say x(size(100x1)),y(size(100x1)) and z(size(100x100)) and i use
z1 = interp2(x,y,z,x1,x2)
to find a scalar value that i need for later calculation. I was wondering if there is a more efficient/quicker way to implement this.scatteredInterpolant function could help? I tried unsuccessfully to implement it with scatteredInterpolant

per isakson

### per isakson (view profile)

on 18 Oct 2014
"find a value" &nbsp should I understand one scalar value?
Dimitrios

### Dimitrios (view profile)

on 18 Oct 2014
thanks for the comment.i edit the question.

### Tags

on 18 Oct 2014

on 18 Oct 2014

scatteredInterpolant is suitable for cases where your source grid is scattered. If you were able to use interp2 it means that your source grid is not scattered. Although you can still use scattered interpolant, it does not give you any additional benefit.
Here are two approaches:
1) using gridded interpolant:
[XGrid,YGrid]=meshgrid(x,y);
F=griddedInterpolant(XGrid',YGrid',z','cubic');
Then you can query that for any point as follow:
z1=F(x1,x2);
x1,x2 could be scattered points or gridded. If it is gridded it should follow NDGrid format.
2) Using Scattered Interpolant:
[XGrid,YGrid]=meshgrid(x,y);
F=scatteredInterpolant(XGrid(:), YGrid(:),z(:));
Then you can query that for any point as follow:
z1=F(x1,x2);

Dimitrios

### Dimitrios (view profile)

on 18 Oct 2014
Tried both approaches.I just changed the cubic to spline.Output value is almost similar(offcurse i didnt expect exactly the same result).I just post the results of computational time:
Interpolation: 5.18sec
griddenInterpolant: 4.8sec
scatteredInterpolant: 10.1sec

### per isakson (view profile)

on 18 Oct 2014
Edited by per isakson

### per isakson (view profile)

on 18 Oct 2014

If linear is good enough as you indicate in the question (default of interp2).
My approach to a similar performance problem included
• avoid the function call. The call itself takes longer than the computation (of a scalar)
• a few lines of code, which do the interpolation, directly in the "calling" code.