Currently the transformations are recalculated for each trial if applied, slowing down training. Instead the possible transformations should be calculated initially and called when selected. This will increase overhead calculations but reduce training time
Currently the transformations are recalculated for each trial if applied, slowing down training. Instead the possible transformations should be calculated initially and called when selected. This will increase overhead calculations but reduce training time
https://github.com/owenodriscoll/AutoML/blob/b2a83c5760903e85721d9e6cfdf98cae2fe4590d/AutoML/automl.py#L263-L268