GitHub user AnthonyTruchet opened a pull request:
https://github.com/apache/spark/pull/16078
[SPARK-18471][MLLIB] Fix huge vectors of zero send in closure in L-BFGS
## What changes were proposed in this pull request?
Introduced util TreeAggregatorWithZeroGenerator and used it to avoid
sending huge zero vector in L-BFGS or similar aggregation.
## How was this patch tested?
Run custom L-BGFS using this aggregator instead of treeAggregate with
significantly increased performance and stability.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/criteo-forks/spark ENG-17719-wrapper
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/16078.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #16078
----
commit b4333a3b89362952e900acd7824e5d40500e3b9f
Author: Anthony Truchet <a.truc...@criteo.com>
Date: 2016-11-29T18:20:38Z
[SPARK-18471][MLLIB] Fix huge vectors of zero send in closure in L-BFGS
Introduced util tTreeAggregatoreWithZeroGenerator to avoid sending huge
zero vector in L-BFGS or similar agregation, as only the size of the zero
value to be generated is captured in the closure.
----
---
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
