You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by tengpeng <gi...@git.apache.org> on 2017/11/05 22:49:18 UTC
[GitHub] spark pull request #19660: [SPARK-18755][WIP][ML] Add Randomized Grid Search...
GitHub user tengpeng opened a pull request:
https://github.com/apache/spark/pull/19660
[SPARK-18755][WIP][ML] Add Randomized Grid Search to Spark ML
## What changes were proposed in this pull request?
Python sklearn has a randomized grid search for reducing the time for parameter tuning:
1. If the candidate parameter values are all discrete, sampling with replacement.
2. If at least one candidate parameter is continuous, sampling without replacement.
This patch mimic the behavior of case 1 only. If we want to do 2, we need significant changes in `ParamGridBuilder` and other cross validation components. This requires more discussions. Thoughts?
Reference: http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.RandomizedSearchCV.html
## How was this patch tested?
Existing test + a new unit test.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/tengpeng/spark CV
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/19660.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 #19660
----
commit 46086202f14185b18351f50c9c09f0641af4bb4f
Author: test <jo...@quetica.com>
Date: 2017-11-05T22:40:51Z
Initial commit for searchRatio
----
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark pull request #19660: [SPARK-18755][WIP][ML] Add Randomized Grid Search...
Posted by tengpeng <gi...@git.apache.org>.
Github user tengpeng closed the pull request at:
https://github.com/apache/spark/pull/19660
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark issue #19660: [SPARK-18755][WIP][ML] Add Randomized Grid Search to Spa...
Posted by srowen <gi...@git.apache.org>.
Github user srowen commented on the issue:
https://github.com/apache/spark/pull/19660
I am not sure this approach adds much. Just lets you shuffle and sample the existing grid. Random hyperparam search is really about sampling a distribution for each hyperparam.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark issue #19660: [SPARK-18755][WIP][ML] Add Randomized Grid Search to Spa...
Posted by AmplabJenkins <gi...@git.apache.org>.
Github user AmplabJenkins commented on the issue:
https://github.com/apache/spark/pull/19660
Can one of the admins verify this patch?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
[GitHub] spark issue #19660: [SPARK-18755][WIP][ML] Add Randomized Grid Search to Spa...
Posted by tengpeng <gi...@git.apache.org>.
Github user tengpeng commented on the issue:
https://github.com/apache/spark/pull/19660
@srowen You are absolutely right. That's the what 2 aims to accomplish. I believe implementing 1 & 2 is the goal, like what they did in sklearn. Need some discussions on design.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org