diff --git a/python/docs/pyspark.streaming.rst b/python/docs/pyspark.streaming.rst
index 5024d694b668f52fa67ad354b9621604d67cece9..f08185627d0bc58ef67344a501762ee2ab92a3de 100644
--- a/python/docs/pyspark.streaming.rst
+++ b/python/docs/pyspark.streaming.rst
@@ -1,5 +1,5 @@
 pyspark.streaming module
-==================
+========================
 
 Module contents
 ---------------
diff --git a/python/pyspark/mllib/__init__.py b/python/pyspark/mllib/__init__.py
index 5030a655fcbba14eadb5814a897815013b567af4..c3217620e3c4ee3005a7912828bdf2da94a5483b 100644
--- a/python/pyspark/mllib/__init__.py
+++ b/python/pyspark/mllib/__init__.py
@@ -32,29 +32,4 @@ import sys
 import rand as random
 random.__name__ = 'random'
 random.RandomRDDs.__module__ = __name__ + '.random'
-
-
-class RandomModuleHook(object):
-    """
-    Hook to import pyspark.mllib.random
-    """
-    fullname = __name__ + '.random'
-
-    def find_module(self, name, path=None):
-        # skip all other modules
-        if not name.startswith(self.fullname):
-            return
-        return self
-
-    def load_module(self, name):
-        if name == self.fullname:
-            return random
-
-        cname = name.rsplit('.', 1)[-1]
-        try:
-            return getattr(random, cname)
-        except AttributeError:
-            raise ImportError
-
-
-sys.meta_path.append(RandomModuleHook())
+sys.modules[__name__ + '.random'] = random
diff --git a/python/pyspark/mllib/feature.py b/python/pyspark/mllib/feature.py
index 741c630cbd6eba32e4a8079e3e985d41610fa072..e46af208866a2a9ffea78d57920d3c1d6b5d636a 100644
--- a/python/pyspark/mllib/feature.py
+++ b/python/pyspark/mllib/feature.py
@@ -53,10 +53,10 @@ class Normalizer(VectorTransformer):
     """
     :: Experimental ::
 
-    Normalizes samples individually to unit L\ :sup:`p`\ norm
+    Normalizes samples individually to unit L\ :sup:`p`\  norm
 
-    For any 1 <= `p` <= float('inf'), normalizes samples using
-    sum(abs(vector). :sup:`p`) :sup:`(1/p)` as norm.
+    For any 1 <= `p` < float('inf'), normalizes samples using
+    sum(abs(vector) :sup:`p`) :sup:`(1/p)` as norm.
 
     For `p` = float('inf'), max(abs(vector)) will be used as norm for normalization.