Regrouping pages on model tuning in the DataOps User Guide#2014
Regrouping pages on model tuning in the DataOps User Guide#2014emassoulie wants to merge 1 commit intomainfrom
Conversation
|
Marked as draft because, in addition to the restructuring, the pages also need to be trimmed a little (with the larger examples perhaps moved to their own section). |
rcap107
left a comment
rcap107
left a comment
There was a problem hiding this comment.
Thanks for the PR @emassoulie, overall I think it's an improvement over the current documentation. I left a couple of comments where I was not convinced by the wording, but aside from that I think we can merge this soon.
|
|
||
| Here are the different kinds of choices, along with their default outcome when | ||
| we are not using hyperparameter search: | ||
| Skrub provides over 10 different ``choose`` methods for tuning use cases, all detailed |
There was a problem hiding this comment.
I wouldn't say there are 10 "different" choose methods: the choose methods are 4 (from, int, float and bool), but they can be used in different ways
|
|
||
| Splitting the data in train and test sets | ||
| ========================================= | ||
| More advanced train/test splitting |
There was a problem hiding this comment.
I think the original title should be kept here: this section isn't about a more advanced way of defining train and test splits, we really are just splitting the data in two
2 participants
The five pages on validation and tuning at the end of the DataOps section of the User Guide have been fused into three main parts:
These parts have examples that should be shortened (or made into "example" pages), and subsection titles that have been made more explicit.