8000 Add table / dictionary for good parameter values · Issue #5004 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

Add table / dictionary for good parameter values #5004

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
amueller opened this issue Jul 20, 2015 · 13 comments
Open

Add table / dictionary for good parameter values #5004

amueller opened this issue Jul 20, 2015 · 13 comments
Labels
Documentation

Comments

@amueller
Copy link
Member

I feel we should add good grids to search over for all estimators, both in the docs and as a dict.
I think a good source would be https://github.com/automl/auto-sklearn
and possibly carret.
@jnothman
Copy link
Member
jnothman commented Aug 3, 2015

I think this would be very handy for newcomers to an algoritgm. Also, I think the process of enumerating these will be revealing in itself, and remind us (and our users) that some parameters are dependent on sample size or number of features... In general we should remind users that "good grids" are applicable to e.g. centred, scaled features.

@kastnerkyle
Copy link
Member

Agreed, nice defaults could be a big benefit

On Sun, Aug 2, 2015 at 9:47 PM, jnothman notifications@github.com wrote:

I think this would be very handy for newcomers to an algoritgm. Also, I
think the process of enumerating these will be revealing in itself, and
remind us (and our users) that some parameters are dependent on sample size
or number of features... In general we should remind users that "good
grids" are applicable to e.g. centred, scaled features.

—
Reply to this email directly or view it on GitHub
#5004 (comment)
.

@GaelVaroquaux
Copy link
Member

👍

@mfeurer
Copy link
Contributor
mfeurer commented Aug 5, 2015

Another source of hyperparameters and their values might be mlr

@amueller
Copy link
Member Author

ping @NoonienSoong

@amueller
Copy link
Member Author

I think the actual package for @mfeurer's parameters is here:
https://bitbucket.org/mfeurer/paramsklearn

See for example here:
https://bitbucket.org/mfeurer/paramsklearn/src/8c4f3368db1e8e65bc123ea94cc2bdac6b7b5faa/ParamSklearn/components/classification/?at=master

@amueller
Copy link
Member Author
amueller commented Sep 9, 2015

@NoonienSoong are you still interested in working on this?

@amueller
Copy link
Member Author
amueller commented Oct 3, 2015

skll could also be a good resource: https://github.com/EducationalTestingService/skll
here:
https://skll.readthedocs.org/en/latest/run_experiment.html#param-grids-optional

@amueller
Copy link
Member Author

maybe also at https://github.com/hyperopt/hyperopt-sklearn
There are some here:
https://github.com/hyperopt/hyperopt-sklearn/blob/master/hpsklearn/components.py
These are distributions, however, and they are probably more than one wants to search over.

@GaelVaroquaux GaelVaroquaux mentioned this issue Oct 23, 2015
Closed
8 tasks
@zermelozf
Copy link
Contributor

I think 's probably best if wait until #4294 is merged before adding the dict (in _constants.py as suggested by @rvraghav93).

@raghavrv
Copy link
Member

From discussions IRL, I feel _constants.py won't be added... You might want to start your work on some foo_bar.py file and after #4294 gets merged, move it into the model_selection module... (and delete the foo_bar.py file)

@amueller
Copy link
Member Author

+1 on what @rvraghav93 said. git moving the file later is no problem.

@zermelozf zermelozf mentioned this issue Oct 23, 2015
Closed
@mfeurer
Copy link
Contributor
mfeurer commented Dec 15, 2015

I just saw that the reference to ParamSklearn from @amueller on the 19th of August points to my old repository; the up-to-date version is in my groups repository https://github.com/automl/paramsklearn/

@amueller amueller mentioned this issue Oct 5, 2018
Closed
@amueller amueller mentioned this issue Jan 22, 2019
Open
@amueller amueller mentioned this issue Jan 7, 2021
Open
@amueller amueller mentioned this issue Nov 30, 2021
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
Documentation
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
9 participants
@jnothman @GaelVaroquaux @amueller @glouppe @zermelozf @kastnerkyle @mfeurer @raghavrv @TomDLT
0