These are chat archives for thunder-project/thunder
MassUnivariateClassifier.fit()returns a Series with the index set to the string
'performance'. However, if multiple features are used, the underlying RDD will have an array of performance values. I'm thinking it would be best of set the index to be a list of lists, with each element denoting the features used (i.e. the
featureSetvariable that is passed to the
fitfunction). Sound legit?
Series.percentile(), all unit tests pass with the new
Series.indexgetter/setter methods. Any future code that tries to create a
Serieswith a wacky index should cause an error. I'll push the final changes for the setter with those corrections to the PR and it should be good to go.