Github user Lewuathe commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-71821238
@mengxr Thank you for reminding me. I create another PR as soon as possible.
---
If your project is set up for it, you can reply to this email and have your
reply appear
Github user asfgit closed the pull request at:
https://github.com/apache/spark/pull/3933
---
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 enab
Github user mengxr commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-71796333
@Lewuathe Do you mind closing this PR? If you want to update the official
document, we can do that in another PR. Thanks for the discussion!
---
If your project is set up
Github user mengxr commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69292824
@Lewuathe For binary classification, the documentation is here:
https://spark.apache.org/docs/latest/mllib-linear-methods.html#evaluation-metrics
But we
Github user Lewuathe commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69273220
@mengxr @srowen Thank you for reviewing. I agree to the reason why `pr`
method is also reasonable in terms of drawing curve. I'll keep it as-is.
But anyway I want to
Github user mengxr commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69240738
@Lewuathe I think we can either deprecate `pr`/`roc` and create
`prCurve`/`rocCurve` with exactly the same behavior, or simply leave them
as-is. I prefer the former a lit
Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69238882
@mengxr That's a fair point, since it's always drawn as a curve with recall
on the x axis and of course the x component is conventionally first. It's a
precision-as-functi
Github user mengxr commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69238136
@srowen By definition, the precision-recall curve consists of (recall,
precision) points. It is unfortunate that it was named this way. I don't have
strong preference on w
Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69181021
@Lewuathe FWIW I vote for deprecating existing methods, but leaving the
behavior the same, and introducing new methods like `prCurve` whose return
value is 'corrected'.
Github user Lewuathe commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69166579
@mengxr I think it is difficult to infer what the order of precision and
recall is returned from `pr` method name. So I think keeping the order of
returned values on met
Github user mengxr commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69066043
@Lewuathe `precision-recall` is widely used. It could be a little strange
to call it "RP". If you feel the method is still confusing even with the doc,
maybe we can add `p
Github user AmplabJenkins commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69021140
Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/25
Github user SparkQA commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69021134
[Test build #25162 has
finished](https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/25162/consoleFull)
for PR 3933 at commit
[`3d38647`](https://gith
Github user SparkQA commented on the pull request:
https://github.com/apache/spark/pull/3933#issuecomment-69014618
[Test build #25162 has
started](https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/25162/consoleFull)
for PR 3933 at commit
[`3d38647`](https://githu
GitHub user Lewuathe opened a pull request:
https://github.com/apache/spark/pull/3933
[SPARK-4284] BinaryClassificationMetrics precision-recall method names s...
...hould correspond to return types
Precision-recall curve is the general term. Therefore I decided not to
chang
15 matches
Mail list logo