search

rapidsai / cuml

cuML - RAPIDS Machine Learning Library
https://docs.rapids.ai/api/cuml/stable/
Apache License 2.0
4.18k stars 527 forks source link

[FEA] Provide support for ensemble.AdaBoost* weighted classifiers #2435

Open teju85 opened 4 years ago

teju85 commented 4 years ago

Is your feature request related to a problem? Please describe. This is based on @zachmayer's feature request on xgboost here. I believe cuML is better suited to provide such an interface.

Describe the solution you'd like

  1. Update all models to do training based on sample-weights
  2. Update all models to support predict-proba (classification cases)
  3. Provide cuml.ensemble.AdaBoost* classes in parity with their sklearn counterparts. (atleast based on cupy functions, to begin with)

Describe alternatives you've considered The alternative is to call sklearn's functions, but I suspect the runtime to be bad due to frequent cpu<->gpu transfers.

Additional context Tagging @tunguz, JFYI.

zachmayer commented 4 years ago

I can already use sklearn.ensemble.AdaBoostClassifier so I don't really need this feature in cuml. I'd love to be able to specify an arbitrary base learner in xgboost, and take advantage of all the cool innovations xgboost brought to boosting.

github-actions[bot] commented 3 years ago

This issue has been marked rotten due to no recent activity in the past 90d. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.

github-actions[bot] commented 3 years ago

This issue has been marked stale due to no recent activity in the past 30d. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be marked rotten if there is no activity in the next 60d.

zachmayer commented 3 years ago

I'd still like this feature!

tunguz commented 3 years ago

Yes, I'd very much like to have this feature.

joaogui1 commented 3 years ago

This would be pretty useful!

tunguz commented 3 years ago

I would still like to see this feature ...