How to pass customized arguments to the fit() function of the learner

[chungdz]

Hi guys, I want to apply catboost as the regressor learner as there're categorical features in the x_cols to fit. I need to identify which columns are categorical data. And the argument to identify them, cat_features, is in the fit() function of the catboost. I just want to know how to pass this argument to the fit() function of the learner? I try to find positions where DoubleMLPLR calls fit() function of the learner but I think they're hidden in the code file deeply such that I still didn't find the positions.

[SvenKlaassen]

Thank you for your interest in the package.
Just to clarify. Isn't it possible use the cat_features argument in CatBoostRegressor (documentation)?

This should be possible to use in DoubleMLPLR

from catboost import CatBoostRegressor
# Initialize data
cat_features = [0, 1]
train_data = [["a", "b", 1, 4, 5, 6],
              ["a", "b", 4, 5, 6, 7],
              ["c", "d", 30, 40, 50, 60]]
train_labels = [1.2, 1.1, -1.3]
eval_data = [["a", "b", 2, 4, 6, 8],
             ["a", "d", 1, 4, 50, 60]]

# Initialize CatBoostRegressor
model = CatBoostRegressor(iterations=2,
                          learning_rate=1,
                          depth=2,
                          cat_features=cat_features)
# Fit model
model.fit(train_data, train_labels)
# Get predicted values
preds = model.predict(eval_data)

If this is not what you are looking for a minimal reproducable example would be helpful.

You can also create your own learners (add additional arguments etc.) as long as they fulfill the minimum requirements.

In case you still want to check where fit() is used:
The predictions of the DoubleMLPLR model are generated here

doubleml-for-py/doubleml/double_ml_plr.py

Line 176 in df617fd

l_hat = _dml_cv_predict(self._learner['ml_l'], x, y, smpls=smpls, n_jobs=n_jobs_cv,

which mostly uses cross_val_predict from scikit-learn

doubleml-for-py/doubleml/_utils.py

Line 63 in df617fd

preds = cross_val_predict(clone(estimator), x, y, cv=smpls, n_jobs=n_jobs, method=method)