[
https://issues.apache.org/jira/browse/FLINK-2157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15252073#comment-15252073
]
ASF GitHub Bot commented on FLINK-2157:
---------------------------------------
Github user thvasilo commented on the pull request:
https://github.com/apache/flink/pull/1849#issuecomment-212977748
I did some testing and I think the problem has to do with the types that
each scaler expects.
`StandardScaler` has fit and transform operations for `DataSets` of type
`Vector`, `LabeledVector`, and `(T :< Vector, Double)` while `MinMaxScaler`
does not provide one for `(T :< Vector, Double)`. If you add the operations the
code runs fine (at least re. you first comment).
So this is a bug unrelated to this PR I think. The question becomes if we
want to support all three of these types. My recommendation would be to have
support for `Vector` and `LabeledVector` only, and remove all operations that
work on `(Vector, Double)` tuples. I will file a JIRA for that.
There is an argument to be whether some pre-processing steps are supervised
(e.g. [PCA vs.
LDA](https://stats.stackexchange.com/questions/161362/supervised-dimensionality-reduction))
but in the strict definition of a transformer we shouldn't care about the
label, only the features, so that operation can implemented at the
`Transformer` level.
> Create evaluation framework for ML library
> ------------------------------------------
>
> Key: FLINK-2157
> URL: https://issues.apache.org/jira/browse/FLINK-2157
> Project: Flink
> Issue Type: New Feature
> Components: Machine Learning Library
> Reporter: Till Rohrmann
> Assignee: Theodore Vasiloudis
> Labels: ML
> Fix For: 1.0.0
>
>
> Currently, FlinkML lacks means to evaluate the performance of trained models.
> It would be great to add some {{Evaluators}} which can calculate some score
> based on the information about true and predicted labels. This could also be
> used for the cross validation to choose the right hyper parameters.
> Possible scores could be F score [1], zero-one-loss score, etc.
> Resources
> [1] [http://en.wikipedia.org/wiki/F1_score]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
