Trying to use interpn - converting sample data into ndgrid format

I'm trying to use the function interpn but it requires that I have everything in the ndgrid format. For example, I'm trying to execute this function:

    Vq = interpn(X1,X2,X3,V,Xq1,Xq2,Xq3);

I'm trying interpolate the value of V given three variables (i.e. performing a 3D interpolation). I have X1, X2, and X3 set up:

    X1 = 1:100; X2 = 1:50; X3 = 1:10;

That is, there are 100 points in X1, 50 points in X2, and 10 points in X1. However, my V is composed differently - it is a (100*50*10) by 1 vector, with each value in V corresponding to each value of X1, X2, and x3. In a sense, V looks like

    [v(x1_1, x2_1, x3_1), v(x1_2, x2_1, x3_1), ... v(x1_100, x2_1, x3_1), v(x1_1, x2_2, x3_1), v(x1_2, x2_2, x3_1), ...].

How can I convert V into the ndgrid format?

I know that I can use griddatan but it is awfully slower than interpn.

Answer 1

[Star Strider]

Use the ndgrid (link) function.

[jh2011]

How? ndgrid creates a grid if you put in the ranges of each dimension, but I have all the values flattened out in a vector

[Star Strider]

You can use ‘X1’, ‘X2’, and ‘X3’ as inputs to ndgrid.

‘V is composed differently - it is a (100*50*10) by 1 vector’

‘V’ is a matrix by that definition. Vectors are by definition (Nx1). If it is in fact a vector, use reshape to create it as a matrix.

Answer 2

[Walter Roberson]

reshape(V, 100, 50, 10)

[jh2011]

What if V is computed based on [X1, X2, X3] created NOT in the ndgrid format?

[Walter Roberson]

If V is based on a rectangular grid, then reshape() it into grid form according to how it is laid out, and then if needed permute() it into ndgrid order.