Calculation hangs in 1D

[Pierre-Andre]

Dear authors

I have a problem in using GCVmoothedNSpline for a given dataset (french csv file attached)

from gcvspline import GCVSmoothedNSpline
import numpy as np
tab=np.loadtxt("bugGCV.csv", delimiter=";",skiprows=1)
tab.shape ##(309, 2)
x=tab[:,0]
y=tab[:,1]
GCV_auto = GCVSmoothedNSpline(x, y) ## hangs

bugGCV.csv

Is it a well known limitation of the algorithm ?

Best regards
Pierre-Andre

PS the output of "pip install" in a new venv

pip install  gcvspline
Collecting gcvspline
  Using cached gcvspline-0.5-cp311-cp311-linux_x86_64.whl
Collecting numpy>=1.12
  Using cached numpy-2.1.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (16.3 MB)
Installing collected packages: numpy, gcvspline
Successfully installed gcvspline-0.5 numpy-2.1.2

and my python3 version

Python 3.11.2 (main, Aug 26 2024, 07:20:54) [GCC 12.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.

[charlesll]

Hi,

To my knowledge, this is not a known behavior.

I reproduced the problem on my computer, while my install is fine and the example notebook runs fine...

A clue is that SmoothedNSpline runs fine, so I suspect the algorithm has a problem converging to a good smoothing factor using the GCV mode.

I noticed that your data are very irregularly spaced, with a X space that is increasing more and more as X increases. Maybe this creates a problem for the GCV routine?

In any case, the loophole seems to be on the side of the FORTRAN code and not of the Python code.