A question about splitting the dataset

[hangtingchen]

Hi, Thank you very much for your explanation of covariate shift. I have a question about the dataset split in the jupyter notebook. You wrote that for fold, (train_idx, test_idx) in enumerate(skf.split(XZ, labels)):, which mixed the real training and test data into the new 'training and test' data in my view. However, it should be ensured that you should not train the model using the test data. I think your original attempt was training the model using the large-scale data and few-shot covariate-shift data, then testing on the test data. Maybe further splitting the test data into training and test is a good idea. https://github.com/erlendd/covariate-shift-adaption/blob/master/Supervised%20classification%20by%20covariate%20shift%20adaption.ipynb