Description
Instead of expanding MLlib based on the current class naming scheme (ProblemWithAlgorithm), we should standardize MLlib's interfaces that clearly separate datasets, formulations, algorithms, parameter sets, and models.
Attachments
Issue Links
- contains
-
SPARK-2692 Decision Tree API update
-
- Closed
-
- is related to
-
SPARK-1457 Change APIs for training algorithms to take optimizer as parameter
-
- Resolved
-
- is superceded by
-
SPARK-10817 ML abstraction umbrella
-
- Resolved
-
1.
|
Pipeline and Parameters |
|
Resolved | Xiangrui Meng |
2.
|
Dataset |
|
Resolved | Xiangrui Meng |
3.
|
Standardize MLlib classes for learners, models |
|
Closed | Joseph K. Bradley |
4.
|
Add API for feature attributes |
|
Resolved | Sean R. Owen |
5.
|
Standardize ML Prediction APIs |
|
Resolved | Joseph K. Bradley |
6.
|
Make ML Prediction Developer APIs public |
|
Resolved | Joseph K. Bradley |
7.
|
Stabilize DecisionTree and ensembles APIs |
|
Resolved | Joseph K. Bradley |