Github user thvasilo commented on the issue:
https://github.com/apache/flink/pull/2819
Hello @gaborhermann,
I really like the idea of introducing a `MatrixFactorization` interface
that we can then use for different specialized optimization algorithms. For the
question I'm afraid I can't be of much help, I'll read the relevant paper this
week and get back to you if I have any more comments.
1. Don't know enough about joins to answer this :/
2. For this we would need to test the two solutions you have proposed and
evaluate the performance. You have listed a couple of pros/cons there, maybe
you can elaborate?
3. If there is absolutely no benefit from using more blocks I don't see the
need to investigate further. We'll need to include these instructions in the
docs.
4. I think test hardening should be done in a different PRs (potentially
for more algorithms). For now manual tests should suffice.
---
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.
---
