DOC add cross-reference to examples instead of duplicating content for GPR#20003
DOC add cross-reference to examples instead of duplicating content for GPR#20003lesteve merged 47 commits intoscikit-learn:mainfrom
Conversation
jjerphan
left a comment
There was a problem hiding this comment.
Thanks for improving those examples @glemaitre. Those are really instructive.
Here are some suggestions from a new-comer to GPR.
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
|
ping @adrinjalali I think the following PR addresses some of your proposals discussed IRL when reviewing #26980 |
|
I like the changes, but in the listed examples, we see: And I think the second and 4th examples don't have a title which is informative enough for users to know what to expect. We should either change the title, or have a text here before referencing them so that the reader knows what they gain by reading the example. |
|
I tried to improve the titles. |
|
@noashin started reviewing this I think, I'll wait for that before doing another review. |
There was a problem hiding this comment.
I would strongly encourage including the term NONPARAMETRIC when discussing GPs.
I would rephrase the opening (lines 9-11):
Gaussian Processes (GP) are a nonparametric supervised learning method used
to solve regression and probabilistic classification problems.
|
|
||
| The disadvantages of Gaussian processes include: | ||
|
|
||
| - They are not sparse, i.e., they use the whole samples/features information to |
There was a problem hiding this comment.
Sparse Gaussian Processes is a thing, just not in Scikit. One defines a set of inducing points (smaller than the data) and use them for learning the GP instead of the data.
This is a good blog post about it.
doc/modules/gaussian_process.rst
Outdated
|
|
||
| The noise level in the targets can be specified by passing it via the parameter | ||
| `alpha`, either globally as a scalar or per datapoint. Note that a moderate | ||
| noise level can also be helpful for dealing with numeric issues during fitting |
There was a problem hiding this comment.
"numerical instabilities" instead of "numerical issues" ?
|
Thanks @noashin. I added the suggestions. |
adrinjalali
left a comment
There was a problem hiding this comment.
Are you gonna do another PR for GPC?
Otherwise LGTM.
| externally for other ways of selecting hyperparameters, e.g., via | ||
| Markov chain Monte Carlo. | ||
|
|
||
| .. topic:: Examples |
There was a problem hiding this comment.
maybe rename to gpr_examples? we might easily conflict with another place defining an Examples topic.
There was a problem hiding this comment.
This is a topic and not the name of the section. I don't think you can use this to reference.
Yes. It would be in the same spirit than what we did for GPR. |
|
No idea why but one Azure build was timing out, since this is a doc PR, I merged it manually. |
…r GPR (scikit-learn#20003) Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
…r GPR (#20003) Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>
…r GPR (scikit-learn#20003) Co-authored-by: Julien Jerphanion <git@jjerphan.xyz>

follow-up #19952
closes #11562
Revamp the documentation of the Gaussian Process user guide.
TODO:
main)