Fitting module updates
Summary of changes
- Added new kwarg to all optimizers
standardize
which is True by default. - Added new
fit.py
which contains a wrapper fit function (which handles the standardization) - The available_fit_methods is now located in
fit.py
rather than inbase_optimizer.py
- The number of nonzero parameters have been added to the
BaseOptimizer
as a property and included insummary
and__str__
- Smaller clean ups here and there
Comments
- The normalization or standardization is done via the methods available in sklearn in their preprocessing module
- In the optimizers the fit matrix and parameters are not changed, its only in the new
fit
method where the scaling and then rescaling happens. - This is done without copying the fit_matrix and hence the fit_matrix goes through a
transform
and then aninverse_transform
which might change it slightly numerically. - Nothing is done in regards to centering the fit matrix or fit data only rescaling of the fit matrix.
Closes #259 (closed) #263 (closed)
Edited by Erik Fransson