MultiOutputClassifier can not work with early_stopping_round?

[yiwiz-sai]

pip3 freeze |grep lightgbm
lightgbm==3.3.2

import lightgbm as lgb
from sklearn.multioutput import*
from sklearn.datasets import * 
params = {'num_leaves':30, 'verbose': -1,
         'num_iterations':10, 'early_stopping_round':3,
         'metric':['auc', 'binary_logloss']}

clf = MultiOutputClassifier(lgb.LGBMClassifier(**params))
# x, y = make_classification(n_classes=2, n_samples=50)
x, y = make_multilabel_classification(n_features=5,n_samples=50, n_classes=5, n_labels=2)
#print(x)
#print(y)
clf.fit(x, y, eval_set=[(x,y)])
clf.predict_proba(x)

get error

ValueError: y should be a 1d array, got an array of shape (50, 5) instead.

I don't find related testing in https://github.com/microsoft/LightGBM/blob/master/tests/python_package_test/test_sklearn.py

[jmoralez]

Hi @yiwiz-sai, thank you for your interest in LightGBM. Indeed this doesn't work because every model takes the eval_set argument as-is and we'd need to do something like eval_set=[(x, y[:, i])] for each model. What MultiOutputClassifier does is just fit a model to each column in y so you could do that manually or use sklearn.ensemble.HistGradientBoostingClassifier which is inspired by LightGBM and builds the validation sets for early stopping internally so that one works in the way you want it to.

Please let us know if this helps.

[yiwiz-sai]

Hi @yiwiz-sai, thank you for your interest in LightGBM. Indeed this doesn't work because every model takes the eval_set argument as-is and we'd need to do something like eval_set=[(x, y[:, i])] for each model. What MultiOutputClassifier does is just fit a model to each column in y so you could do that manually or use sklearn.ensemble.HistGradientBoostingClassifier which is inspired by LightGBM and builds the validation sets for early stopping internally so that one works in the way you want it to.

Please let us know if this helps.

I see, thanks. can lightgbm support this in the future ?

[jmoralez]

I think that feature request is already in #2302 as #3313.

[github-actions]

This issue has been automatically locked since there has not been any recent activity since it was closed.
To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues
including a reference to this.