Closed corochann closed 4 years ago
Use label_scaler.inverse_transform for both train/predict It makes code very easy to understand.
Computational time for always postprocess inverse_transform is quite small and i think this is acceptable.
inverse_transform
no-compat because GraphConvPredictor.__call__ behavior will change to apply inverse_transform.
GraphConvPredictor.__call__
Use label_scaler.inverse_transform for both train/predict It makes code very easy to understand.
Computational time for always postprocess
inverse_transform
is quite small and i think this is acceptable.