Github user dbtsai commented on the pull request:
https://github.com/apache/spark/pull/11610#issuecomment-197213572
I'm not an expert in this area, but after thinking it more, I don't think
we can use `DGELSD` which minimizes `||b - A*x||` using the singular value
decomposition (SVD) of a rectangular matrix `A`, which may be rank-deficient.
Since `A` is `n x m` matrix where `n` is the number of training samples, `A^T
A` is trackable locally. Unless we have distributed version of `DGELSD`; this
will not work. Seems that eigen decomposition is the good approach.
For those least square problems, when the number of columns is small, we
can always solve it by `A^TA x = A^Tb`. Why do we need tall-skinny QR/SVD for
stability? Isn't it for the case that the number of columns is larger?
Thanks.
---
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]
