05-clarity

[ltetrel]

@manojneuro Thanks for your submission.
This issue will review notebook 05.

2. Circular Inference: How to avoid double dipping

The GridSearchCV method that you will learn about below makes it easy (though not guaranteed) to avoid double dipping. In previous exercises we examined cases where double dipping is clear (e.g., training on all of the data and testing on a subset). However, double dipping can be a lot more subtle and hard to detect, for example in situations where you perform feature selection on the entire dataset before classification (as in last week's notebook).

You introduce the term "double dipping" without explaining it. I think you could add few words to explain it ?

3.2 Regularization Example: L2 vs. L1

I think the figure could gain more clarity if it was 3x3, and by labeling the axis ("-th fold" for example)

[manojneuro]

@ltetrel: We did cover double dipping in the previous notebook, but we can provide references etc. on the topic here.

For the L2 vs. L1, we can enhance the figure and the description to make it clearer.

[ltetrel]

Thanks @manojneuro