Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Conversation

@pat-s
Copy link
Member

@pat-s pat-s commented Oct 8, 2019

closes #2642

  • Update default for param shrinkage from 0.001 to 0.1
  • add param n.cores and suppress parallelization by default
  • update param distribution with the actual package defaults
  • add some more choices for param distribution

@pat-s pat-s requested review from larskotthoff and mllg October 8, 2019 12:25
@pat-s pat-s marked this pull request as ready for review October 8, 2019 12:25
note = "`keep.data` is set to FALSE to reduce memory requirements. Note on param 'distribution': gbm will select 'bernoulli' by default for 2 classes, and 'multinomial' for multiclass problems. The latter is the only setting that works for > 2 classes. It is recommended to NOT use the 'multinomial' setting, see https://github.com/mlr-org/mlr/issues/2612",
note = paste0(collapse = "", c("`keep.data` is set to FALSE to reduce memory requirements.\n",
"Note on param 'distribution': gbm will select 'bernoulli' by default for 2 classes, and 'multinomial' for multiclass problems.\n",
"The latter is the only setting that works for > 2 classes. It is recommended to NOT use the 'multinomial' setting, see https://github.com/mlr-org/mlr/issues/2612.",
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is confusing. There's no "multinomial" value anymore, and the default doesn't depend on the task here.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, good catch. Removed the note about "multinomial" entirely as it does not apply anymore.

name = "Gradient Boosting Machine",
short.name = "gbm",
note = '`keep.data` is set to FALSE to reduce memory requirements, `distribution` has been set to `"gaussian"` by default.',
note = poaste0(collapse = "", c('`keep.data` is set to FALSE to reduce memory requirements, `distribution` has been set to `"gaussian"` by default.',
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Mentioned default value for distribution is incorrect. Same note on distribution as for classification learner here?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah I was lazy here. Sorry. Default now "gaussian" and the classif learner was changed as well.

@pat-s pat-s merged commit 5564bfb into master Oct 13, 2019
@pat-s pat-s deleted the gbm branch October 13, 2019 16:52
vrodriguezf pushed a commit to vrodriguezf/mlr that referenced this pull request Jan 16, 2021
* upd gbm distribution argument

* upd shrinkage and distribtion argument

* update learner note

* document n.cores change

* add NEWS

* correct regr.gbm distribution param

* clear note about 'multinomial' param for classif.gbm

* fix classif_gbm test

* add multionomial option for distribution param again

* Deploy from Travis build 14223 [ci skip]

Build URL: https://travis-ci.org/mlr-org/mlr/builds/597271702
Commit: 2e4bceb
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

regr.gbm default shrinkage

3 participants