I am using ICET to predict energies I get from DFT calculations. However, my cross validation error is really high. I have played with different variables and still it is very high. I want to know if I am doing something wrong or anything I could do to decrease the error.
Its hard to say without more information, it could related to choice of cutoffs and how many training structures used.
What do you get when printing the optimizer object,
This is the snapshot of my code but my error is beyond reasonable. I wanted to know if my code has some error.
Looks like your energies are very large.
Are the energies you use per atom?
For debugging I would recommend using least-squares and see if you can first get the training error lower.
My energies were not per atom which created the issue