[SPARK-29464][PYTHON][ML] PySpark ML should expose Params.clear() to unset a user supplied Param

### What changes were proposed in this pull request?
change PySpark ml ```Params._clear``` to ```Params.clear```

### Why are the changes needed?
PySpark ML currently has a private _clear() method that will unset a param. This should be made public to match the Scala API and give users a way to unset a user supplied param.

### Does this PR introduce any user-facing change?
Yes. PySpark ml ```Params._clear``` ---> ```Params.clear```

### How was this patch tested?
Add test.

Closes #26130 from huaxingao/spark-29464.

Authored-by: Huaxin Gao <[email protected]>
Signed-off-by: Bryan Cutler <[email protected]>

Author: huaxingao

python/pyspark/ml/classification.py

@@ -446,7 +446,7 @@ def setThreshold(self, value):
         Clears value of :py:attr:`thresholds` if it has been set.
         """
         self._set(threshold=value)
-        self._clear(self.thresholds)
+        self.clear(self.thresholds)
         return self
 
     @since("1.4.0")
@@ -477,7 +477,7 @@ def setThresholds(self, value):
         Clears value of :py:attr:`threshold` if it has been set.
         """
         self._set(thresholds=value)
-        self._clear(self.threshold)
+        self.clear(self.threshold)
         return self
 
     @since("1.5.0")

python/pyspark/ml/param/__init__.py

@@ -452,7 +452,7 @@ def _set(self, **kwargs):
             self._paramMap[p] = value
         return self
 
-    def _clear(self, param):
+    def clear(self, param):
         """
         Clears a param from the param map if it has been explicitly set.
         """

python/pyspark/ml/tests/test_param.py

@@ -27,8 +27,8 @@
 from pyspark.ml.classification import LogisticRegression
 from pyspark.ml.clustering import KMeans
 from pyspark.ml.feature import Binarizer, Bucketizer, ElementwiseProduct, IndexToString, \
-    VectorSlicer, Word2Vec
-from pyspark.ml.linalg import DenseVector, SparseVector
+    MaxAbsScaler, VectorSlicer, Word2Vec
+from pyspark.ml.linalg import DenseVector, SparseVector, Vectors
 from pyspark.ml.param import Param, Params, TypeConverters
 from pyspark.ml.param.shared import HasInputCol, HasMaxIter, HasSeed
 from pyspark.ml.wrapper import JavaParams
@@ -224,6 +224,10 @@ def test_params(self):
         testParams.setMaxIter(100)
         self.assertTrue(testParams.isSet(maxIter))
         self.assertEqual(testParams.getMaxIter(), 100)
+        testParams.clear(maxIter)
+        self.assertFalse(testParams.isSet(maxIter))
+        self.assertEqual(testParams.getMaxIter(), 10)
+        testParams.setMaxIter(100)
 
         self.assertTrue(testParams.hasParam(inputCol.name))
         self.assertFalse(testParams.hasDefault(inputCol))
@@ -248,6 +252,18 @@ def test_params(self):
                        "maxIter: max number of iterations (>= 0). (default: 10, current: 100)",
                        "seed: random seed. (default: 41, current: 43)"]))
 
+    def test_clear_param(self):
+        df = self.spark.createDataFrame([(Vectors.dense([1.0]),), (Vectors.dense([2.0]),)], ["a"])
+        maScaler = MaxAbsScaler(inputCol="a", outputCol="scaled")
+        model = maScaler.fit(df)
+        self.assertTrue(model.isSet(model.outputCol))
+        self.assertEqual(model.getOutputCol(), "scaled")
+        model.clear(model.outputCol)
+        self.assertFalse(model.isSet(model.outputCol))
+        self.assertEqual(model.getOutputCol()[:12], 'MaxAbsScaler')
+        output = model.transform(df)
+        self.assertEqual(model.getOutputCol(), output.schema.names[1])
+
     def test_kmeans_param(self):
         algo = KMeans()
         self.assertEqual(algo.getInitMode(), "k-means||")

python/pyspark/ml/wrapper.py

@@ -280,6 +280,14 @@ def copy(self, extra=None):
             that._transfer_params_to_java()
         return that
 
+    def clear(self, param):
+        """
+        Clears a param from the param map if it has been explicitly set.
+        """
+        super(JavaParams, self).clear(param)
+        java_param = self._java_obj.getParam(param.name)
+        self._java_obj.clear(java_param)
+
 
 @inherit_doc
 class JavaEstimator(JavaParams, Estimator):

python/pyspark/testing/mlutils.py

@@ -62,7 +62,7 @@ def check_params(test_self, py_stage, check_params_exist=True):
                 continue  # Random seeds between Spark and PySpark are different
             java_default = _java2py(test_self.sc,
                                     java_stage.clear(java_param).getOrDefault(java_param))
-            py_stage._clear(p)
+            py_stage.clear(p)
             py_default = py_stage.getOrDefault(p)
             # equality test for NaN is always False
             if isinstance(java_default, float) and np.isnan(java_default):