Merge pull request #11 from ml-cube/dev-new-proposition

docs + fix: new proposition, minor fixes

Author: Giovanni Giacometti

README.md

@@ -1,26 +1,30 @@
 import logging
 
 import numpy as np
-from ml3_drift.sklearn.univariate.ks import KSDriftDetector
+from ml3_drift.monitoring.algorithms.batch.ks import KSAlgorithm
 from sklearn.tree import DecisionTreeRegressor
 from sklearn.pipeline import Pipeline
 from sklearn.preprocessing import StandardScaler
 from ml3_drift.callbacks.base import logger_callback
 from functools import partial
 
+from ml3_drift.sklearn.base import SklearnDriftDetector
+
 logger = logging.getLogger(__name__)
 
 
 if __name__ == "__main__":
     # Define your pipeline as usual, but also add a drift detector
-    drift_detector = KSDriftDetector(
-        callbacks=[
-            partial(
-                logger_callback,
-                logger=logger,
-                level=logging.CRITICAL,
-            )
-        ]
+    drift_detector = SklearnDriftDetector(
+        KSAlgorithm(
+            callbacks=[
+                partial(
+                    logger_callback,
+                    logger=logger,
+                    level=logging.CRITICAL,
+                )
+            ]
+        )
     )
 
     pipeline = Pipeline(

examples/sklearn/continuous_data.py

@@ -1,14 +1,15 @@
 from functools import partial
 import logging
-from ml3_drift.sklearn.univariate.ks import KSDriftDetector
-from ml3_drift.sklearn.univariate.chi_square import ChiSquareDriftDetector
+from ml3_drift.monitoring.algorithms.batch.ks import KSAlgorithm
+from ml3_drift.monitoring.algorithms.batch.chi_square import ChiSquareAlgorithm
 
 from sklearn.tree import DecisionTreeRegressor
 from sklearn.pipeline import Pipeline
 from sklearn.compose import ColumnTransformer
 from sklearn.preprocessing import StandardScaler, OrdinalEncoder
 import numpy as np
 from ml3_drift.callbacks.base import logger_callback
+from ml3_drift.sklearn.base import SklearnDriftDetector
 
 
 logger = logging.getLogger(__name__)
@@ -51,27 +52,31 @@
         transformers=[
             (
                 "cont",
-                KSDriftDetector(
-                    callbacks=[
-                        partial(
-                            logger_callback,
-                            logger=logger,
-                            level=logging.CRITICAL,
-                        ),
-                    ]
+                SklearnDriftDetector(
+                    KSAlgorithm(
+                        callbacks=[
+                            partial(
+                                logger_callback,
+                                logger=logger,
+                                level=logging.CRITICAL,
+                            ),
+                        ]
+                    )
                 ),
                 [0, 1],
             ),
             (
                 "cat",
-                ChiSquareDriftDetector(
-                    callbacks=[
-                        partial(
-                            logger_callback,
-                            logger=logger,
-                            level=logging.CRITICAL,
-                        ),
-                    ]
+                SklearnDriftDetector(
+                    ChiSquareAlgorithm(
+                        callbacks=[
+                            partial(
+                                logger_callback,
+                                logger=logger,
+                                level=logging.CRITICAL,
+                            ),
+                        ]
+                    )
                 ),
                 [2, 3],
             ),

examples/sklearn/mixed_data_monitoring.py

@@ -4,3 +4,6 @@ def __init__(
     ):
         self.concepts = concepts
         self.same_distributions = same_distributions
+
+    def __repr__(self):
+        return f"Report(concepts={self.concepts}, same_distributions={dict(self.same_distributions)})"

src/ml3_drift/analysis/report.py

@@ -1,9 +1,11 @@
+import dataclasses
 from logging import Logger
-from ml3_drift.callbacks.models import DriftInfo
+
+from ml3_drift.models.monitoring import DriftInfo
 
 
 def logger_callback(
-    drift_info: DriftInfo,
+    drift_info: DriftInfo | None,
     logger: Logger,
     level: int,
 ) -> None:
@@ -20,14 +22,24 @@ def logger_callback(
     """
 
     if drift_info is None:
-        logger.log(level, "Drift Detected!")
+        logger.log(level, "Drift Detected, no drift info provided!")
         return
-    msg = f"Drift detected on feature at index {drift_info.feature_index} by drift detector {drift_info.drift_detector}."
 
-    if drift_info.p_value is not None:
-        msg += f"\n p-value = {drift_info.p_value}"
+    logger.log(level, f"Drift Detected, drift info: {dataclasses.asdict(drift_info)}")
+
+
+def print_callback(drift_info: DriftInfo | None) -> None:
+    """
+    Print callback prints a message to the console when drift is detected.
+    It should be used only for testing purposes.
+
+    Example
+    -------
 
-    if drift_info.threshold is not None:
-        msg += f"\n Threshold = {drift_info.threshold}"
+    callback = print_callback
+    """
+    if drift_info is None:
+        print("Drift Detected, no drift info provided!")
+        return
 
-    logger.log(level, msg)
+    print(f"Drift Detected, drift info: {dataclasses.asdict(drift_info)}")

src/ml3_drift/callbacks/base.py

@@ -48,9 +48,8 @@ def __init__(
         self.seed = seed
         self._args = args
         self._kwargs = kwargs
-        super().__init__(
-            comparison_size=1, callbacks=callbacks
-        )  # since we add only one sample per step and river handles building the window internally we set comparison_size to 1
+        # since we add only one sample per step and river handles building the window internally we set comparison_size to 1
+        super().__init__(comparison_size=1, callbacks=callbacks)
 
     def _reset_internal_parameters(self):
         self.drift_agent = RiverKSWIN(

src/ml3_drift/callbacks/models.py

@@ -40,6 +40,7 @@ def test_kswin_univariate_one_drift(self, abrupt_univariate_online_drift_info):
         )
 
     def test_kswin_univariate_two_drift(self, abrupt_univariate_online_bidrift_info):
+        np.random.seed(42)
         data_stream, drift_point_1, drift_point_2 = (
             abrupt_univariate_online_bidrift_info
         )