>"If they included feature generation in the training process and ran it once per fold, it would be OK" How so? The data used as validation in one fold would b

>"If they included feature generation in the training process and ran it once per fold, it would be OK" How so? The data used as validation in one fold would be used to determine features in the next...