Matlab interpolate (interpn) ignore dimensions with NaN [duplicate]

StackOverflow https://stackoverflow.com/questions/12457446

Question

Possible Duplicate:
MATLAB: Using interpolation to replace missing values (NaN)

I would like to interpolate data in multiple dimensions using the interpn command.

The problem is that the data contains NaN's. Therefore, in some dimensions, there are less than 2 datapoints which prevents matlab from interpolating. I would like interpn to return a NaN value in that case.

EXAMPLE: My data looks somehow likes this 

  val(:,:,1) =
    -2.3810   -0.9365
     NaN   -0.6910


  val(:,:,2) =
    -2.3354   -0.9271
    NaN   -0.6859

Matlab gives an error when interpolating since there are inly NaN's in one dimension. I would like matlab to ignore this error and instead return NaN's form the interpn function.

Was it helpful?

Solution

You can introduce a go-between function which looks for NaNs.

So instead of calling interpn, you can call a function that does the following:

searches for NaN via any(any(any(isnan(val))) if true returns NaN, otherwise it calls your function.

It's not clear if you want NaN for the entire interpn, or just for a particular layer/dimension. If it's the latter, I'd appreciate it if you showed a bit more code.

Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top