Home

ToDo:

For Estimators:

• implement assurances that any variable listed in hyper_params is also a property of the class
• convert set_* methods to Dependent methods

For crossValidators

• convert set_* methods to Dependent properties
• convert fold_lbls into Dependent property

For extendability:

• Check nomenclature of scikit learn matlab ports and make things as consistent as possible (to make it as easy as possible to convert code from other git repos now and down the line)

Implement GridSearchCV

Implement nonlinearSVM (Regression and Classification)

Implement calibratedClf

Implement MLPCR Modify regression scorers so that they fall back on yfit_raw if yfit is unavailable (with a warning)

review source reconstruction code, and implement checks for linear models (check paper to ensure linear models are valid)

multiclassLinearSvmClf needs decisionFcn update. Estimator needs an update to throw a warning when something like decisionFcn is likely to happen.

Scorers may need some work. You need to be able to make the following happen

• generalize @get_hinge_loss to multiclass scenario. scikit-learn may have ideas: https://github.com/scikit-learn/scikit-learn/blob/95119c13a/sklearn/metrics/_classification.py#L2284
• Update binary classifiers (esp. svmClf) to incorporate arbitrary class labels, but still work with hinge loss.