[python-package] How can I use goss in the latest version 4.5 about lightgbm HOT 4 CLOSED

By the way, I reformatted your follow-up question. If you're unsure how I did that, please see https://docs.github.com/en/get-started/writing-on-github/getting-started-with-writing-and-formatting-on-github/basic-writing-and-formatting-syntax.

from lightgbm.

Thanks for using LightGBM.

As described at https://lightgbm.readthedocs.io/en/latest/Parameters.html#data_sample_strategy, GOSS can be specified via data_sample_strategy. That change was made in LightGBM #4827, as part of v4.0.0 (https://github.com/microsoft/LightGBM/releases/tag/v4.0.0).

Here's a minimal, reproducible example:

import lightgbm as lgb
from sklearn.datasets import make_regression

X, y = make_regression(n_samples=10_000, n_features=10)
reg = lgb.LGBMRegressor(
  n_estimators=10,
  data_sample_strategy="goss"
)
reg.fit(X, y)

If you do that, you'll see this in the logs:

[LightGBM] [Info] Using GOSS

from lightgbm.

got it! Thanks a lot.Now I have a new question:
In the doc LightGBM regressor includes following params:

__init__(boosting_type='gbdt', num_leaves=31, max_depth=-1, learning_rate=0.1, n_estimators=100, subsample_for_bin=200000, objective=None, class_weight=None, min_split_gain=0.0, min_child_weight=0.001, min_child_samples=20, subsample=1.0, subsample_freq=0, colsample_bytree=1.0, reg_alpha=0.0, reg_lambda=0.0, random_state=None, n_jobs=None, importance_type='split', **kwargs)

But no data_sample_strategy inside.

Why can it be directly changed?

from lightgbm.

But no 'data_sample_strategy' inside.Why can it be directly changed?

Any other parameters passed to the constructor of LGBMRegressor which don't match its keyword arguments will be collected together and sent to LightGBM's C++ library as additional parameters.

As it says at https://lightgbm.readthedocs.io/en/latest/pythonapi/lightgbm.LGBMRegressor.html:

**kwargs –

Other parameters for the model. Check http://lightgbm.readthedocs.io/en/latest/Parameters.html for more parameters.

from lightgbm.