Closed yamsgithub closed 7 years ago
You are correct. I'll release a new version that conforms to the semantics.
Based on request from the D3M folks, we can split our imputation into multiple classes. Please, use the following classes: GreedyImputation, IterativeRegressionImputation, MICE, KNNImputation, MeanImputation
Thanks @kyao.
@yamsgithub the parameter greater_is_better
was used in the old all-in-one imputer. Now it is separated into different classes and conformed with unified interface, so no need for greater_is_better
anymore.
@yamsgithub Just pushed changes with the correct get_params() and set_params(). Also, uploaded to PyPI dsbox-datacleaning-0.2.1. Please, let me know if the changes work for you.
Great! thanks!
Both Encoder and Imputation seem to be setting and getting hyperparameters instead of the learned parameters which is the semantics as defined by the unified interface.
Also I do not see the parameter greater_is_better being used anywhere (it is only mentioned in the doc string comments) in Imputation.