Parallelize approximation of gradients (Vadere)
Parallelize approximate_gradient function in VadereModel.
At the moment, only the evaluation of the parameter directions at one sample point are run in parallel
option 1:
collect all parameter configurations (as a dict) and pass them all to the suq-controller.
this means for each sample point and all uncertain parameters (finite difference approximation needs to be calculated in each parameter direction) - forward (+h) and backward (-h) for central differences
option 2:
parallelize python code in approximate_gradient and call the suq-controller several times in parallel.
parallel structure for function calls can be seen in request.py from suq-controller (uses package multiprocessing).
compare _mp_query.