Emulating in fortran the trained neural network in PES-learn

[Bingsnow]

So i extracted the parameters(weights and bias ) of the trained model from Pes learn and have written a Fortran code (implementing the scaling technique which was used in the compute.energy.py ) .when i compare the results from pes learn python function and fortran there is a slight differnce in innermost coordinates and the rest of the output is same .Could you please tell me ,where is the issue ? below i have attached the graph comparing both the result

[sgoodlett]

It is difficult to say where this difference might be coming from with this information. I might suggest plotting your data set on the same graph to at least see which plot is more correct?