Replacing numpy repmat with tile

NumPy matlib has been deprecated. Replacing the usage accordingly.

Signed-off-by: Patrick Bloebaum <[email protected]>

dowhy/gcm/anomaly.py

@@ -2,7 +2,6 @@
 
 import numpy as np
 import pandas as pd
-from numpy.matlib import repmat
 from tqdm import tqdm
 
 from dowhy.gcm import config
@@ -40,7 +39,9 @@ def conditional_anomaly_scores(
 
     result = np.zeros(parent_samples.shape[0])
     for i in range(parent_samples.shape[0]):
-        samples_from_conditional = causal_mechanism.draw_samples(repmat(parent_samples[i], num_samples_conditional, 1))
+        samples_from_conditional = causal_mechanism.draw_samples(
+            np.tile(parent_samples[i], (num_samples_conditional, 1))
+        )
         anomaly_scorer = anomaly_scorer_factory()
         anomaly_scorer.fit(samples_from_conditional)
         result[i] = anomaly_scorer.score(target_samples[i])[0]

dowhy/gcm/distribution_change.py

@@ -6,7 +6,6 @@
 import networkx as nx
 import numpy as np
 import pandas as pd
-from numpy.matlib import repmat
 from statsmodels.stats.multitest import multipletests
 from tqdm import tqdm
 
@@ -496,7 +495,7 @@ def _estimate_distribution_change_score(
     for joint_parent_sample in joint_parent_samples:
         old_result = result
 
-        samples = repmat(joint_parent_sample, num_joint_samples, 1)
+        samples = np.tile(joint_parent_sample, (num_joint_samples, 1))
         result += difference_estimation_func(
             causal_model_original.draw_samples(samples), causal_model_new.draw_samples(samples)
         )

dowhy/gcm/influence.py

@@ -7,7 +7,6 @@
 import numpy as np
 import pandas as pd
 from joblib import Parallel, delayed
-from numpy.matlib import repmat
 
 import dowhy.gcm.auto as auto
 from dowhy.gcm import feature_relevance_sample
@@ -175,7 +174,7 @@ def _estimate_direct_strength(
     average_difference_result = 0
     converged_run = 0
     for run, sample in enumerate(distribution_samples):
-        tmp_samples = repmat(sample, num_samples_conditional, 1)
+        tmp_samples = np.tile(sample, (num_samples_conditional, 1))
 
         rnd_permutation = np.random.choice(distribution_samples.shape[0], num_samples_conditional, replace=False)
 
@@ -436,7 +435,9 @@ def icc_set_function(subset: np.ndarray) -> Union[np.ndarray, float]:
             # In case of the empty subset (all are jointly randomize), it boils down to taking the average over all
             # predictions, seeing that the randomization yields the same values for each sample of interest (none of the
             # samples of interest are used to replace a (jointly) 'randomized' sample).
-            predictions = repmat(np.mean(prediction_method(noise_samples), axis=0), baseline_noise_samples.shape[0], 1)
+            predictions = np.tile(
+                np.mean(prediction_method(noise_samples), axis=0), (baseline_noise_samples.shape[0], 1)
+            )
         else:
             predictions = marginal_expectation(
                 prediction_method,

dowhy/gcm/model_evaluation.py

@@ -6,7 +6,6 @@
 import numpy as np
 import pandas as pd
 from joblib import Parallel, delayed
-from numpy.matlib import repmat
 from scipy.stats import mode
 from sklearn.metrics import f1_score, mean_squared_error, r2_score
 from sklearn.model_selection import KFold
@@ -730,7 +729,7 @@ def empirical_crps(generated_Y, observed_y):
         all_classes = np.unique(Y)
 
         for x, y in zip(X, Y):
-            samples = conditional_sampling_method(repmat(x, num_conditional_samples, 1))
+            samples = conditional_sampling_method(np.tile(x, (num_conditional_samples, 1)))
 
             sample_categorical_crps = []
             for cat in all_classes:
@@ -749,7 +748,7 @@ def empirical_crps(generated_Y, observed_y):
 
         for x, y in zip(X, Y):
             crps_values.append(
-                empirical_crps(conditional_sampling_method(repmat(x, num_conditional_samples, 1)) / std_Y, y / std_Y)
+                empirical_crps(conditional_sampling_method(np.tile(x, (num_conditional_samples, 1))) / std_Y, y / std_Y)
             )
 
     return float(np.mean(np.array(crps_values)))

tests/gcm/test_stats.py

@@ -1,7 +1,6 @@
 import numpy as np
 import pytest
 from flaky import flaky
-from numpy.matlib import repmat
 from pytest import approx
 
 from dowhy.gcm.ml import (
@@ -433,7 +432,7 @@ def test_given_nonlinear_categorical_data_when_evaluate_marginal_expectation_the
 def test_given_different_batch_sizes_when_estimating_marginal_expectation_then_returns_correct_marginal_expectations():
     X = np.random.normal(0, 1, (34, 3))
     feature_samples = np.random.normal(0, 1, (123, 3))
-    expected_non_aggregated = np.array([repmat(X[i, :], feature_samples.shape[0], 1) for i in range(X.shape[0])])
+    expected_non_aggregated = np.array([np.tile(X[i, :], (feature_samples.shape[0], 1)) for i in range(X.shape[0])])
 
     def my_pred_func(X: np.ndarray) -> np.ndarray:
         return X.copy()