Hi, new here, but I have a question that I'm wondering.
I wanted to build some models based on output predictions from other models, but ran into issues when naively putting these estimators into a feature union.
Predictors has no transform method, although in a way you could see them as transforming X into y. Is there a underlying reason why this is not the case?, e.g. predictors could have a transform that returns a column vector instead of an array, and then also other features expected from transformers like the features out etc. Is there anything talking against that predictors can be seen as transformers as well, will we wreak havoc in other places of the code, where predictors could suddenly be mistaken for transformers due to having the expected methods?