Xgboost: incremental learning, partial_fit like sklearn?

[xiaoleihuang]

My data is too big to fit into memory, do xgboost support partial_fit like sklearn? support incremental learning.

[xgdgsc]

I' m also looking for this feature.

[xiaoleihuang]

Hi, @xgdgsc Plz check this one below:
xgboost.train(params, dtrain, num_boost_round=10, evals=(), obj=None, feval=None, maximize=False, early_stopping_rounds=None, evals_result=None, verbose_eval=True, xgb_model=None, callbacks=None, learning_rates=None)

This API solves problem.
The link of documentation is: http://xgboost.readthedocs.io/en/latest/python/python_api.html#xgboost.train

Sorry I forgot to update my issue....

[xgdgsc]

Thanks. That's cool. But adding a similar API as sklearn could also be helpful.

[xiaoleihuang]

@xgdgsc Yeah,the model supports incremental training, you might read the source code there. It won't affect your usage, it's just a different name...

[ylogx]

Here is a gist to demonstrate this.

[nickkaranatsios]

The link to gist file is broken

[antontarasenko]

@xgdgsc

xgboost.XGBClassifier.fit() supports parameter xgb_model for incremental learning:

• Docs: http://xgboost.readthedocs.io/en/latest/python/python_api.html#xgboost.XGBClassifier.fit
• Code: https://github.com/dmlc/xgboost/blob/master/python-package/xgboost/sklearn.py#L217

[ylogx]

@nickkaranatsios Fixed it. Changed my username recently.