pearsonr_mat sometimes throws divide by zero error if simulation finishes without any elution

[ronald-jaepel]

On an optimization I'm running the

def pearsonr_mat(x, Y, times):

sometimes throws an error

  File "/home/IBT/jaepel/miniforge3/envs/CADETProcessGP/lib/python3.10/site-packages/CADETProcess/comparison/difference.py", line 279, in evaluate
    metric = self._evaluate(solution)
  File "/home/IBT/jaepel/miniforge3/envs/CADETProcessGP/lib/python3.10/site-packages/CADETProcess/comparison/difference.py", line 582, in _evaluate
    corr, offset_original = pearson(
  File "/home/IBT/jaepel/miniforge3/envs/CADETProcessGP/lib/python3.10/site-packages/CADETProcess/comparison/shape.py", line 194, in pearson
    pearson = eval_offsets(
  File "/home/IBT/jaepel/miniforge3/envs/CADETProcessGP/lib/python3.10/site-packages/CADETProcess/comparison/shape.py", line 129, in eval_offsets
    scores = pearsonr_mat(reference_data, rol_mat, time)
ZeroDivisionError: division by zero

When debugging, I can see that the simulation in this case was only fluctuations around 1e-13. I didn't think to save the sim results and it takes ages to get back to that point in the optimization (~2h) so I'll export a dataset for easier debugging next week.

[ronald-jaepel]

Closed by #155