Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/3975#issuecomment-69335485
On the one hand, this changes behavior, and causes the result to not quite
match what's in the input file. Some parts of MLlib handle -1 / +1 for binary
classification, although generally 0 / 1 is expected. That said, this method
already changes what it reads by modifying 1-based feature indices to 0-based.
FWIW I think this would be a good change since -1 / +1 is sometimes used in
libsvm, and -1 does not always work within MLlib. Some docs and a test to
accompany this would be appropriate. You might also search the code base for
instances where this translation already happens and remove it if needed, but I
didn't see such a thing.
(And the JIRA / PR title could be updated to reflect what the issue really
is now)
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
