Fix leaderboard in estimators.py

[lamauamal]

Fixed an issue in the leaderboard functionality in autosklearn/estimators.py.

Issue

When setting the details to be displayed on the leaderboard, the "data_preprocessor" field remains empty. Upon checking the code, it’s clear that the actual key used is "data_preprocessing", which doesn’t match the "data_preprocessor" key used in runhistory.json. And this issue cannot be resolved by directly modifying "data_preprocessing" to "data_preprocessor", because there’s another entry in runhistory.json, "data_preprocessor:feature_type:numerical_transformer:rescaling:__choice__", which also meets the filtering criteria of if “data_preprocessor” in key and “__choice__” in key.

• example:

import sklearn.metrics
import autosklearn.classification
from sklearn.datasets import load_iris
import autosklearn.classification

X, y = load_iris(return_X_y=True)
X_train, X_test, y_train, y_test = sklearn.model_selection.train_test_split(
    X, y, random_state=1
)

automl = autosklearn.classification.AutoSklearnClassifier(
    time_left_for_this_task=120,
    per_run_time_limit=30,
    include={
            "classifier": ["decision_tree", "mlp"]
        },
)
automl.fit(X_train, y_train)
print(automl.leaderboard(detailed=True).head(2).to_string())

# Before:
#           rank  ensemble_weight           type  cost  duration  config_id  train_loss  seed    start_time      end_time  budget              status data_preprocessors    feature_preprocessors balancing_strategy           config_origin
# model_id
# 22           4             0.02            mlp   0.0  2.240190         21    0.026667     0  1.779353e+09  1.779353e+09     0.0  StatusType.SUCCESS                 []  [feature_agglomeration]          weighting  Random Search (sorted)
# 29           1             0.04  decision_tree   0.0  0.604004         28    0.026667     0  1.779353e+09  1.779353e+09     0.0  StatusType.SUCCESS                 []             [polynomial]               none           Random Search

# After:
#           rank  ensemble_weight           type  cost  duration  config_id  train_loss  seed    start_time      end_time  budget              status data_preprocessors    feature_preprocessors balancing_strategy           config_origin
# model_id
# 22           5             0.02            mlp   0.0  0.839216         21    0.026667     0  1.779353e+09  1.779353e+09     0.0  StatusType.SUCCESS     [feature_type]  [feature_agglomeration]          weighting  Random Search (sorted)
# 29           3             0.08  decision_tree   0.0  0.576417         28    0.026667     0  1.779353e+09  1.779353e+09     0.0  StatusType.SUCCESS     [feature_type]             [polynomial]               none           Random Search

Changes line 1154 in estimators.py

• Before:

"data_preprocessors": [
                        value
                        for key, value in run_config.items()
                        if "data_preprocessing" in key and "__choice__" in key
                    ]

• After:
  "data_preprocessors": [run_config.get("data_preprocessor:__choice__")]