Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Incomplete
-
None
-
None
Description
As discussed in SPARK-14489 and the related PR, additional strategies may be useful in the case of unknown users/items at prediction time for ALS. These include:
- some form of "average rating" as a default
- computing predictions using the average user/item vector
There may be other interesting approaches too. This ticket tracks the addition of these more advanced strategies.
Attachments
Issue Links
- is related to
-
SPARK-14489 RegressionEvaluator returns NaN for ALS in Spark ml
- Resolved