I'm trying to use HalvingGridSearch to find the best DecisionTree model. My model performs a multi-label prediction on a single example, it is trained on a batch of data of size (n_samples x n_features=9418) and has an output of size (n_samples x n_output_features=281). The output labels contain multiple classes for a given example and use one-hot encoding, so the vector is in this form: [0,0,1,0,1,...,1]. tree_classifier = DecisionTreeClassifier(random_state=42) train_split = int(tree_train_fraction * len(labels)) tree_inputs_train = tree_inputs[:train_split] tree_inputs_eval = tree_inputs[train_split:] labels_train = labels[:train_split] labels_eval = labels[train_split:] param_grid = { 'max_depth': [ 2, 4, 6, 8, None], 'min_samples_split': [2, 5, 10], 'min_samples_leaf': [1, 2, 3, 4] } search = HalvingGridSearchCV( estimator = tree_classifier, param_grid = param_grid, factor = 2, # ogni iterazione dimezza il numero di candidati resource = 'n_samples', max_resources = 'auto', scoring = 'f1_micro', random_state = 42 ) search.fit(tree_inputs_train, labels_train) tree_classifier = search.best_estimator_ it return the following error when attempting to fit the models: ValueError: y should be a 1d array, got an array of shape (154, 281) instead. I've tried fitting the same models using GridSearchCV and it doesn't return this error. Is it possible that there is a bug in the HalvingGridSearchCV implementation? full error traceback: Traceback (most recent call last): File "/home/gbenanti/Tesi_Benanti/tesi-magistrale-marchetti/DoctorXAI/statistics.py", line 828, in fidelity_synth, hit_synth, fidelity_synth_variance, hit_synth_variance = main(interval, ontological_perturbation, generative_perturbation) File "/home/gbenanti/Tesi_Benanti/tesi-magistrale-marchetti/DoctorXAI/statistics.py", line 801, in main tree, tree_inputs_encoded, model_labels_encoded = explain_label(synt_neigh_icd, synt_neigh_counts, neigh_labels_onehot, max_icd_id, tree_train_fraction, File "/home/gbenanti/Tesi_Benanti/tesi-magistrale-marchetti/DoctorXAI/statistics.py", line 232, in explain_label search.fit(tree_inputs_train, labels_train) File "/home/gbenanti/.venv/lib/python3.10/site-packages/sklearn/base.py", line 1389, in wrapper return fit_method(estimator, *args, **kwargs) File "/home/gbenanti/.venv/lib/python3.10/site-packages/sklearn/model_selection/_search_successive_halving.py", line 247, in fit self._check_input_parameters( File "/home/gbenanti/.venv/lib/python3.10/site-packages/sklearn/model_selection/_search_successive_halving.py", line 164, in _check_input_parameters y = validate_data(self, X="no_validation", y=y) File "/home/gbenanti/.venv/lib/python3.10/site-packages/sklearn/utils/validation.py", line 2946, in validate_data out = _check_y(y, **check_params) File "/home/gbenanti/.venv/lib/python3.10/site-packages/sklearn/utils/validation.py", line 1408, in _check_y y = column_or_1d(y, warn=True) File "/home/gbenanti/.venv/lib/python3.10/site-packages/sklearn/utils/validation.py", line 1485, in column_or_1d raise ValueError( ValueError: y should be a 1d array, got an array of shape (153, 281) instead. Continue reading...
