Github user thvasilo commented on the pull request:
https://github.com/apache/flink/pull/700#issuecomment-117053448
Hello @peedeeX21 . The API does not deal with distributed models at the
moment. In the K-means case having the model distributed is overkill, as it is
highly unlikely that you will have >1000 centroids, making the model tiny, and
distributing it actually creates unnecessary overhead.
We can keep the current implementation, but in the future we should really
test against a non distributed model, which can be broadcast in a
DataSet[Seq[LabeledVector]] and compare performance.
Also, could you add an evaluate operation (EvaluateDataSetOperation) for
Kmeans (and corresponding test)? It would be parametrized as
EvaluateDataSetOperation[Kmeans, Vector, Double]
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
