Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/12299#issuecomment-208873482
So if the input is vector A, then instead of computing A'A we're computing,
for some mean vector M, (A-M)'(A-M) = A'A - A'M - M'A + M'M. A'A remains
efficient to calculate and M'M can be precomputed by the caller and passed in.
A'M and M'A are transposes and can be computed once, and that can be done
somewhat efficiently since A is sparse. It's more complex for sure; I wasn't
sure whether for the sizes we're contemplating, whether just pushing one dense
vector to BLAS would still be faster.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
