LGTM. I'd really like to see related projects, vis a vis projects with a compatible/similar API, becoming a more freely managed catalogue to emphasise that scikit-learn is (in one of its manifestations) an ecosystem. Any ideas how best to achieve that?

Options:

• Github wiki page. Pros: anyone can edit it, so there can be more contributions. Cons: less moderation, links can become broken. Discoverability may be solved by adding a prominent link from scikit-learn docs.
• something like "awesome-sklearn" repository, similar to zillions of other "awesome-.." repositories. Pros/cons: ? I'm not an user of such repositories, but maybe it can be easier for people to subscribe for updates by watching the repository. With wiki or doc page you need to check it from time to time in order to find new related packages. Cons: 'awesome-...' repository name is lame.
• improve existing "related projects" page: add a call for adding more projects right at http://scikit-learn.org/stable/related_projects.html page, with instructions (please send a PR if you know a project with an API similar/compatible to sklearn and you want it to be listed...). Add a more prominent link to this page. Pros: moderation. Cons: ?

something like "awesome-sklearn" repository

Maybe in scikit-learn-contrib we could add a repository "related-projects" with a README filled with such links?

@raghavrv Having it in scikit-learn-contrib organization makes sense: not all packages can be moved to scikit-learn-contrib organization even if they satisfy all conditions - e.g. a company may want to keep a repo under its name, or someone may disagree with giving commit and admin access to a third party even if it is a highly trusted third party.

Indeed. I'm suggesting just for the links. (Instead of a new org/repository 'related-projects' or 'other-projects).

I'm suggesting so because scikit-learn-contrib is now a sizeable collection of "related projects"... Adding a repository there with a readme linking all other related projects that may not be moved there would complete it as a collection of all projects / API wrappers related to sklearn...

Other thing you've subtly pointed out to me is that we don't want this to
be constrained to /stable/ so at a minimum we need a separate CI process to
publish changes to it. (or use github wiki)

On 15 November 2016 at 22:50, Mikhail Korobov notifications@github.com
wrote:

Options:

• Github wiki page. Pros: anyone can edit it, so there can be more
  contributions. Cons: less moderation, links can become broken.
  Discoverability may be solved by adding a prominent link from scikit-learn
  docs.
• something like "awesome-sklearn" repository, similar to zillions of
  other "awesome-.." repositories. Pros/cons: ? I'm not an user of such
  packages, but maybe it can be easier for people to subscribe for updates by
  watching the repository. With wiki or doc page you need to check it from
  time to time in order to find new related packages. Cons: 'awesome-...'
  repository name is lame.
• improve existing "related projects" page: add a call for adding more
  projects right at http://scikit-learn.org/stable/related_projects.html
  page, with instructions (please send a PR if you know a project with an API
  similar/compatible to sklearn and you want it to be listed...). Add a more
  prominent link to this page. Pros: moderation. Cons: ?

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#7878 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAEz622TbtXBunUtvo1TNtypea5Z1XPaks5q-ZyJgaJpZM4Kyacd
.

Is it too unfettered to let the wiki store a YaML listing that could then
have an update hook that copies it to the web server (with checks and
balances to ensure it's reasonable size and format) for display/search with
a DataTable or similar?

On 16 November 2016 at 08:07, Joel Nothman joel.nothman@gmail.com wrote:

Other thing you've subtly pointed out to me is that we don't want this to
be constrained to /stable/ so at a minimum we need a separate CI process to
publish changes to it. (or use github wiki)

On 15 November 2016 at 22:50, Mikhail Korobov notifications@github.com
wrote:

Options:

• Github wiki page. Pros: anyone can edit it, so there can be more
  contributions. Cons: less moderation, links can become broken.
  Discoverability may be solved by adding a prominent link from scikit-learn
  docs.
• something like "awesome-sklearn" repository, similar to zillions of
  other "awesome-.." repositories. Pros/cons: ? I'm not an user of such
  packages, but maybe it can be easier for people to subscribe for updates by
  watching the repository. With wiki or doc page you need to check it from
  time to time in order to find new related packages. Cons: 'awesome-...'
  repository name is lame.
• improve existing "related projects" page: add a call for adding
  more projects right at http://scikit-learn.org/stable
  /related_projects.html page, with instructions (please send a PR if
  you know a project with an API similar/compatible to sklearn and you want
  it to be listed...). Add a more prominent link to this page. Pros:
  moderation. Cons: ?

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#7878 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAEz622TbtXBunUtvo1TNtypea5Z1XPaks5q-ZyJgaJpZM4Kyacd
.

Github wiki page

We do have this page.

to emphasise that scikit-learn is (in one of its manifestations) an ecosystem

Yes and so scikit-learn should focus more on solving the hard problems (model selection, sample weights, etc).

If we put this into scikit-learn-contrib we don't need an additional CI process.
And I like the idea to put it there. I'm not sure about additional sub-repo vs just putting it into https://github.com/scikit-learn-contrib/scikit-learn-contrib (and maybe creating a website for that)

👍 to merge this now, add the "related projects" to "scikit-learn-contrib" and then just link to there from here.

I'm not sure about additional sub-repo

The advantage of a dedicated repo either under scikit-learn or scikit-learn-contrib is that it reaches a broader audience by forks and stars. Making a PR to add their project is also much easier. You can see most of the awesome-* repos are quite useful in presenting all related stuff together in a nice page like this one for all machine learning related stuff and this one for tensorflow. You won't even be needing a dedicated webpage per se...

I am actually wondering if we should simply name it awesome-sklearn. I can understand that its quite lame but it seems to be quite popular with that name. And as in above links we could also link all the (sci/*)py(data/con) talks/tutorials, books and nice projects (not just libraries) related to scikit learn or the scikit API.

The advantage of a dedicated repo either under scikit-learn or scikit-learn-contrib is that it reaches a broader audience by forks and stars.

I don't understand. Why is that easier for a related-projects repo than for the scikit-learn-contrib repo?

The contrib is for projects that are hosted in that repo. The related-projects? would be about anything and everything related to sklearn. (Books / Popular blogs / Kaggle pages (maybe?) and ofcourse other projects too...)... These things get croud sourced pretty soon...

I was expecting scikit-contrib itself to be one such link inside that repo...

Yeah but why does that need a separate repo instead of a section in the readme? that seems overkill

Yeah but why does that need a separate repo instead of a section in the readme

I'd rather not clutter the README file with non-contrib projects. And if we start referencing non-contrib projects, what's the benefit of becoming a contrib project?

I don't feel that we receive so many related-project PRs that we need to rethink our workflow.

👍 to merge this now, add the "related projects" to "scikit-learn-contrib" and then just link to there from here.

I'm merging this for now. We can have a separate discussion on this if needed...

Thanks @kmike