Transform Estimators (GLRM, Aggregator) cannot be used in the sklearn.pipeline properly due to transform property

Description

The transform estimators should implement transform and fit method to be used in sklearn.pipeline. H2OGeneralizedLowRankEstimator, and H2OAggregatorEstimator have @property transform which sklearn.pipeline incorrectly used to calculate transformation data. It caused TypeError.

Note: It depends on the order the estimator is placed in a pipeline - sometimes transform method is not used in pipeline and no error is thrown.

See this PR how this problem was solved in H2OPrincipalComponentAnalysisEstimator and in H2OSingularValueDecompositionEstimator.

The same approach should be used in H2OSingularValueDecompositionEstimator, but for H2OGeneralizedLowRankEstimator and H2OAggregatorEstimator should be implemented another approach, for example, a new special subclass for the usage in the sklearn.pipeline which implements transform and fit method, but no properties.

Related to (Initialization of H2OSingularValueDecompositionEstimator does not work now).

Assignee

Veronika Maurerová

Fix versions

Reporter

Veronika Maurerová

Support ticket URL

None

Labels

Affected Spark version

None

Customer Request Type

None

Task progress

None

CustomerVisible

No

Priority

Major
Configure