search

LMZimmer / Auto-PyTorch_refactor

Apache License 2.0
0 stars 2 forks source link

Use train/val/test data for the encoding to prevent not seen classes #12

Closed franchuterivera closed 3 years ago

franchuterivera commented 4 years ago

This line only uses the training data to build an encoding space: https://github.com/LMZimmer/Auto-PyTorch_refactor/blob/720bde8d873a272c5a29ddb2fd742a7affdcbd71/autoPyTorch/pipeline/components/preprocessing/encoding/OrdinalEncoder.py#L25

To prevent issues, employ train, test, val data IF available to build and fit encoders.

ravinkohli commented 3 years ago

I think this issue can be closed.