RxRx1 Data Augmentation

[niki-amini-naieni]

Hello,

It seems as though RxRx1 already includes data augmentation for the training data for all baseline methods as shown below

(https://github.com/p-lambda/wilds/blob/main/examples/transforms.py)

Could you please confirm that this is the case and provide the reasoning for this decision? Why was this done for RxRx1 and not for some other datasets?

Thank you!

[kohpangwei]

Yes, it does include data augmentation. For each dataset, we picked some default augmentations based on prior work, and this augmentation was what was used in our reference implementation for RxRx1.

[niki-amini-naieni]

Great, thanks for the confirmation!