erroranalysis 0.5.2__tar.gz → 0.5.4__tar.gz

erroranalysis-0.5.4/MANIFEST.in

@@ -0,0 +1,2 @@
+include requirements.txt
+include requirements-object-detection.txt

{erroranalysis-0.5.2/erroranalysis.egg-info → erroranalysis-0.5.4}/PKG-INFO

@@ -1,26 +1,33 @@
 Metadata-Version: 2.1
 Name: erroranalysis
-Version: 0.5.2
+Version: 0.5.4
 Summary: Core error analysis APIs
 Home-page: https://github.com/microsoft/responsible-ai-widgets
 Author: Roman Lutz, Ilya Matiach, Ke Xu
 Author-email: raiwidgets-maintain@microsoft.com
-Classifier: Programming Language :: Python :: 3.6
 Classifier: Programming Language :: Python :: 3.7
 Classifier: Programming Language :: Python :: 3.8
 Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
 Classifier: License :: OSI Approved :: MIT License
 Classifier: Operating System :: OS Independent
 Classifier: Development Status :: 3 - Alpha
-Requires-Python: >=3.6
+Requires-Python: >=3.7
 Description-Content-Type: text/markdown
-Provides-Extra: object_detection
 License-File: LICENSE
+Requires-Dist: numpy>=1.17.2
+Requires-Dist: pandas<2.0.0,>=0.25.1
+Requires-Dist: scipy>=1.4.1
+Requires-Dist: scikit-learn>=0.22.1
+Requires-Dist: lightgbm>=2.0.11
+Requires-Dist: raiutils>=0.4.2
+Provides-Extra: object-detection
+Requires-Dist: vision_explanation_methods; extra == "object-detection"
 
 # Error Analysis SDK for Python
 
-### This package has been tested with Python 3.6, 3.7, 3.8, 3.9 and 3.10
+### This package has been tested with Python 3.7, 3.8, 3.9, 3.10 and 3.11
 
 The error analysis sdk enables users to get a deeper understanding of machine learning model errors. When evaluating a machine learning model, aggregate accuracy is not sufficient and single-score evaluation may hide important conditions of inaccuracies. Use Error Analysis to identify cohorts with higher error rates and diagnose the root causes behind these errors.
 

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/README.md

@@ -1,6 +1,6 @@
 # Error Analysis SDK for Python
 
-### This package has been tested with Python 3.6, 3.7, 3.8, 3.9 and 3.10
+### This package has been tested with Python 3.7, 3.8, 3.9, 3.10 and 3.11
 
 The error analysis sdk enables users to get a deeper understanding of machine learning model errors. When evaluating a machine learning model, aggregate accuracy is not sufficient and single-score evaluation may hide important conditions of inaccuracies. Use Error Analysis to identify cohorts with higher error rates and diagnose the root causes behind these errors.
 

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/erroranalysis/_internal/process_categoricals.py

@@ -49,5 +49,5 @@ def process_categoricals(all_feature_names: List[str],
             categories.append(category_values)
             category_dictionary[category_index] = category_values
 
-    return categories, categorical_indexes, category_dictionary,\
+    return categories, categorical_indexes, category_dictionary, \
         string_ind_data

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/erroranalysis/analyzer/error_analyzer.py

@@ -482,7 +482,7 @@ class BaseAnalyzer(ABC):
         except ValueError:
             # Impute input_data if it contains NaNs, infinity or a value too
             # large for dtype('float64')
-            input_data = np.nan_to_num(input_data)
+            input_data = np.nan_to_num(input_data.astype(float))
             importances = self._compute_error_correlation(
                 input_data, diff, error_correlation_method)
         return importances

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/erroranalysis/error_correlation_methods/gbm.py

@@ -12,8 +12,8 @@ from erroranalysis._internal.constants import ModelTask
 
 def compute_gbm_global_importance(input_data, diff, model_task,
                                   categorical_indexes):
-    """Compute global importance score for EBM between the features and error.
-    :param input_data: The input data to compute the EBM global importance
+    """Compute global importance score for GBM between the features and error.
+    :param input_data: The input data to compute the GBM global importance
         score on.
     :type input_data: numpy.ndarray
     :param diff: The difference between the label and prediction
@@ -21,7 +21,7 @@ def compute_gbm_global_importance(input_data, diff, model_task,
     :type diff: numpy.ndarray
     :param model_task: The model task.
     :type model_task: str
-    :return: The computed EBM global importance score between the features and
+    :return: The computed GBM global importance score between the features and
         error.
     :rtype: list[float]
     """
@@ -33,6 +33,11 @@ def compute_gbm_global_importance(input_data, diff, model_task,
     model.fit(input_data, diff, categorical_feature=categorical_indexes)
     explainer = shap.TreeExplainer(model)
     shap_values = explainer.shap_values(input_data)
+    dims = np.shape(shap_values)
+    # fix some inconsistencies in the shape of the shap_values
+    # for newer versions of shap>=0.45.0 for single-valued target column
+    if is_classification and len(dims) == 2:
+        shap_values = np.expand_dims(shap_values, axis=0)
     shap_mean_abs = np.abs(shap_values).mean(axis=0)
     if is_classification:
         shap_mean_abs = shap_mean_abs.mean(axis=0)

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/erroranalysis/version.py

@@ -4,5 +4,5 @@
 name = 'erroranalysis'
 _major = '0'
 _minor = '5'
-_patch = '2'
+_patch = '4'
 version = '{}.{}.{}'.format(_major, _minor, _patch)

{erroranalysis-0.5.2 → erroranalysis-0.5.4/erroranalysis.egg-info}/PKG-INFO

@@ -1,26 +1,33 @@
 Metadata-Version: 2.1
 Name: erroranalysis
-Version: 0.5.2
+Version: 0.5.4
 Summary: Core error analysis APIs
 Home-page: https://github.com/microsoft/responsible-ai-widgets
 Author: Roman Lutz, Ilya Matiach, Ke Xu
 Author-email: raiwidgets-maintain@microsoft.com
-Classifier: Programming Language :: Python :: 3.6
 Classifier: Programming Language :: Python :: 3.7
 Classifier: Programming Language :: Python :: 3.8
 Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
 Classifier: License :: OSI Approved :: MIT License
 Classifier: Operating System :: OS Independent
 Classifier: Development Status :: 3 - Alpha
-Requires-Python: >=3.6
+Requires-Python: >=3.7
 Description-Content-Type: text/markdown
-Provides-Extra: object_detection
 License-File: LICENSE
+Requires-Dist: numpy>=1.17.2
+Requires-Dist: pandas<2.0.0,>=0.25.1
+Requires-Dist: scipy>=1.4.1
+Requires-Dist: scikit-learn>=0.22.1
+Requires-Dist: lightgbm>=2.0.11
+Requires-Dist: raiutils>=0.4.2
+Provides-Extra: object-detection
+Requires-Dist: vision_explanation_methods; extra == "object-detection"
 
 # Error Analysis SDK for Python
 
-### This package has been tested with Python 3.6, 3.7, 3.8, 3.9 and 3.10
+### This package has been tested with Python 3.7, 3.8, 3.9, 3.10 and 3.11
 
 The error analysis sdk enables users to get a deeper understanding of machine learning model errors. When evaluating a machine learning model, aggregate accuracy is not sufficient and single-score evaluation may hide important conditions of inaccuracies. Use Error Analysis to identify cohorts with higher error rates and diagnose the root causes behind these errors.
 

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/erroranalysis.egg-info/SOURCES.txt

@@ -1,6 +1,7 @@
 LICENSE
 MANIFEST.in
 README.md
+requirements-object-detection.txt
 requirements.txt
 setup.py
 erroranalysis/__init__.py

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/erroranalysis.egg-info/requires.txt

@@ -3,7 +3,7 @@ pandas<2.0.0,>=0.25.1
 scipy>=1.4.1
 scikit-learn>=0.22.1
 lightgbm>=2.0.11
-raiutils>=0.4.0
+raiutils>=0.4.2
 
 [object_detection]
 vision_explanation_methods

erroranalysis-0.5.4/requirements-object-detection.txt

@@ -0,0 +1 @@
+vision_explanation_methods

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/requirements.txt

@@ -3,4 +3,4 @@ pandas>=0.25.1,<2.0.0
 scipy>=1.4.1
 scikit-learn>=0.22.1
 lightgbm>=2.0.11
-raiutils>=0.4.0
+raiutils>=0.4.2

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/setup.py

@@ -30,15 +30,15 @@ setuptools.setup(
     long_description_content_type="text/markdown",
     url="https://github.com/microsoft/responsible-ai-widgets",
     packages=setuptools.find_packages(),
-    python_requires='>=3.6',
+    python_requires='>=3.7',
     install_requires=install_requires,
     extras_require=extras_require,
     classifiers=[
-        "Programming Language :: Python :: 3.6",
         "Programming Language :: Python :: 3.7",
         "Programming Language :: Python :: 3.8",
         "Programming Language :: Python :: 3.9",
         "Programming Language :: Python :: 3.10",
+        "Programming Language :: Python :: 3.11",
         "License :: OSI Approved :: MIT License",
         "Operating System :: OS Independent",
         "Development Status :: 3 - Alpha"

{erroranalysis-0.5.2 → erroranalysis-0.5.4}/tests/test_importances.py

@@ -6,6 +6,7 @@ import time
 import numpy as np
 import pytest
 from common_utils import replicate_dataset
+from sklearn.base import BaseEstimator
 
 from erroranalysis._internal.constants import (ErrorCorrelationMethods,
                                                ModelTask)
@@ -139,6 +140,42 @@ class TestImportances(object):
         scores = model_analyzer.compute_importances(error_correlation_method)
         assert len(scores) == DEFAULT_SAMPLE_COLS
 
+    @pytest.mark.parametrize('num_rows', [3, 4])
+    @pytest.mark.parametrize('nan_correlation_method',
+                             [MUTUAL_INFO, EBM, GBM_SHAP])
+    def test_nan_data_importances(self, num_rows, nan_correlation_method):
+        # validate we can run on very few rows
+        X_train, _, X_test, y_test, _ = \
+            create_binary_classification_dataset(NUM_SAMPLE_ROWS)
+        feature_names = list(X_train.columns)
+
+        class DummyModel(BaseEstimator):
+            def fit(self, X, y=None):
+                return self
+
+            def predict(self, X):
+                return np.zeros((len(X),), dtype=bool)
+
+            def predict_proba(self, X):
+                return np.zeros((len(X),), dtype=bool)
+
+        # Use the dummy model
+        model = DummyModel()
+
+        X_test = X_test[:num_rows]
+        y_test = y_test[:num_rows]
+
+        # Randomly replace some values in X_test with NaN
+        nan_mask = np.random.choice([True, False], size=X_test.shape)
+        X_test = X_test.astype(float)
+        X_test[nan_mask] = np.nan
+
+        categorical_features = []
+        model_analyzer = ModelAnalyzer(model, X_test, y_test,
+                                       feature_names,
+                                       categorical_features)
+        model_analyzer.compute_importances(nan_correlation_method)
+
     @pytest.mark.parametrize('error_correlation_method',
                              [MUTUAL_INFO, EBM, GBM_SHAP])
     def test_importances_missings(self, error_correlation_method):

erroranalysis-0.5.2/MANIFEST.in

@@ -1 +0,0 @@
-include requirements.txt