_get_kriging_matrix is very long for large datasets

[alexisrosuel]

Hi again,

I just saw that the function '_get_kriging_matrix' is very very long (about 5-10 secs) when the training points size is about 10,000 or 15,000.

Do you know if a speed up would be possible ? For instance, I was thinking that as I specified in my model 'n_closed_points' = 100, computing the kriging matrix over the 100 nearest points would be sufficient ?

[rth]

@alexisrosuel The bottleneck must be in the cdist and variogram calculations in _get_kriging_matrix .

Would this issue be related to a local variogram model discussed in #41?

[alexisrosuel]

Yes indeed, I was not sure yesterday, but it is clear now that this is the same issue

[basaks]

@rth Yes, very much so. A local variogram model could help in this situation.
I am still looking to get some funding for that work.

[MuellerSeb]

I would say the problem is, that even with n_closest_points given, the whole matrix is constructed. This should be only done for the points needed.

[MuellerSeb]

This should be discussed in #143. Closing for now. Feel free to re-open or go on with discussion in the linked issue.