interpn vs griddata: how are results different?

Starting off from here:

A=data;
[X,Y,Z]=ndgrid(1:61,1:73,1:61);
x2=linspace(1, 61, 31);
y2=linspace(1, 73, 37);
z2=linspace(1, 61, 31);
[X2 Y2 Z2]=ndgrid(x2, y2, z2);

How are these variables below different? The correlation between data1 and data2 is r=.97.

data1=interpn(X2,Y2,Z2, A, X, Y, Z, 'linear');
data2=griddata(X2,Y2,Z2,A,X,Y,Z,'linear'); 

Thank you!

[Adam Danz]

What about this answer?

https://www.mathworks.com/matlabcentral/answers/517853-interpn-griddata-difference#answer_426034

Answer 1

[Bruno Luong]

Mathematical, in short the difference is

• INTERPN uses tensorial linear function on nd-hypercube that enclose the query point.
• GRIDDATA uses linear function on nd-simplex the query point

[Tahariet Sharon]

Thanks. Practically, what would be the difference?

[Bruno Luong]

INTERPN is faster and provides interpolation result more smooth (C1-discontinuouson smaller set) and better symmetric (invariant if you exchange coordinates) than GRIDDATA. If you have initial point on grid use INTERPN.

Use GRIDDATA only when you can't use the other (scattered data).

[Tahariet Sharon]

Thank you!