Don’t tweak your model after measuring the generalization error: you would just start overfitting the test set.
- B. Machine Learning Project Checklist
- from Hands-On Machine Learning with Scikit-Learn and TensorFlow
- Publisher: O'Reilly Media, Inc.
- Released: March 2017
Not too problematic for k-fold cross-validation (I'm assuming)
Share this highlighthttp://www.safaribooksonline.com/a/hands-on-machine-learning/19621778/