How to set learning rate decay on sklearn interface like LGBMClassifier

[BovenPeng]

Summary

I have searched some relevant information before, but I only get the implement of learning rate decay in LightGBM original interface like follows:
#129
https://lightgbm.readthedocs.io/en/latest/Parameters.html
https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/56158

gbm = lgb.train(
    params,
    lgb_train,
    num_boost_round=10,
    init_model=gbm,
    valid_sets=lgb_eval,
    callbacks=[lgb.reset_parameter(learning_rate = lambda current_round: alpha * e**(k*current_round)
)

And the parameter refit_decay_rate controls the leaf_output, which is kind of like to avoid overfitting.
Sorry that I didn't find some useful relevant information about it so I was not sure of its use.

Motivation

So I want to know there is possible to implement learning rate decay in the sklearn interface, LGBMClassifier(), like the lgb original code example before.
Thanks in advance.

[StrikerRUS]

Hi @BovenPeng !

Sure, you can write learning rate decay for the sklearn wrapper just like for the original interface:

est = lgb.LGBMClassifier().fit(X, y, callbacks=[lgb.reset_parameter(learning_rate=lambda current_round: alpha * e**(k * current_round))])

LightGBM/python-package/lightgbm/sklearn.py

Lines 445 to 447 in dd16fa9

	callbacks : list of callback functions or None, optional (default=None)
	List of callback functions that are applied at each iteration.
	See Callbacks in Python API for more information.

``