[branch-1.1][SPARK-3990] add a note on ALS usage

Because we switched back to Kryo in #3187 , we need to leave a note about the workaround.

Author: Xiangrui Meng <meng@databricks.com>

Closes #3190 from mengxr/SPARK-3990-1.1 and squashes the following commits:

d4818f3 [Xiangrui Meng] fix python style
53725b0 [Xiangrui Meng] add a note about SPARK-3990
56ad70e [Xiangrui Meng] add a note about SPARK-3990

1 files changed, 12 insertions, 0 deletions

diff --git a/python/pyspark/mllib/recommendation.py b/python/pyspark/mllib/recommendation.py
index e863fc249e..d4c06c0566 100644
--- a/python/pyspark/mllib/recommendation.py
+++ b/python/pyspark/mllib/recommendation.py
@@ -60,6 +60,18 @@ class MatrixFactorizationModel(object):
 
 class ALS(object):
 
+    """Alternating Least Squares matrix factorization.
+
+    SPARK-3990: In Spark 1.1.x, we use Kryo serialization by default in
+    PySpark. ALS does not work under this default setting. You can switch
+    back to the default Java serialization by setting:
+
+    spark.serializer=org.apache.spark.serializer.JavaSerializer
+
+    Please go to http://spark.apache.org/docs/latest/configuration.html
+    for instructions on how to configure Spark.
+    """
+
     @classmethod
     def train(cls, ratings, rank, iterations=5, lambda_=0.01, blocks=-1):
         sc = ratings.context