[jira] [Commented] (SPARK-31007) KMeans optimization based on triangle-inequality
[
https://issues.apache.org/jira/browse/SPARK-31007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17489270#comment-17489270
]
zhengruifeng commented on SPARK-31007:
--
this case is not OOM, but the overflow:
{code:java}
scala> val k = 5
val k: Int = 5scala> k * (k + 1) / 2
val res0: Int = -897458648scala> k * (k + 1)
val res1: Int = -1794917296scala> k / 2 * (k + 1)
val res2: Int = 1250025000scala> Array.ofDim[Double](k * (k + 1) / 2)
java.lang.NegativeArraySizeException
at scala.reflect.ManifestFactory$DoubleManifest.newArray(Manifest.scala:273)
at scala.reflect.ManifestFactory$DoubleManifest.newArray(Manifest.scala:271)
at scala.Array$.ofDim(Array.scala:323)
... 32 elidedscala> Array.ofDim[Double](k / 2 * (k + 1))
java.lang.OutOfMemoryError: Java heap space
at scala.reflect.ManifestFactory$DoubleManifest.newArray(Manifest.scala:273)
at scala.reflect.ManifestFactory$DoubleManifest.newArray(Manifest.scala:271)
at scala.Array$.ofDim(Array.scala:323)
... 29 elidedscala> val k = 45000
val k: Int = 45000scala> Array.ofDim[Double](k / 2 * (k + 1))
java.lang.OutOfMemoryError: Java heap spacescala> k / 2 * (k + 1)
val res6: Int = 1012522500
{code}
I think we should add a limit of k for this optimization. the change should not
be large, I will send a PR.
> KMeans optimization based on triangle-inequality
>
>
> Key: SPARK-31007
> URL: https://issues.apache.org/jira/browse/SPARK-31007
> Project: Spark
> Issue Type: Improvement
> Components: ML
>Affects Versions: 3.1.0
>Reporter: zhengruifeng
>Assignee: zhengruifeng
>Priority: Major
> Fix For: 3.1.0
>
> Attachments: ICML03-022.pdf
>
>
> In current impl, following Lemma is used in KMeans:
> 0, Let x be a point, let b be a center and o be the origin, then d(x,c) >=
> |(d(x,o) - d(c,o))| = |norm(x)-norm(c)|
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}
> According to [Using the Triangle Inequality to Accelerate
> K-Means|[https://www.aaai.org/Papers/ICML/2003/ICML03-022.pdf]], we can go
> futher, and there are another two Lemmas can be used:
> 1, Let x be a point, and let b and c be centers. If d(b,c)>=2d(x,b) then
> d(x,c) >= d(x,b);
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}.
> However, luckily for CosineDistance we can get a variant in the space of
> radian/angle.
> 2, Let x be a point, and let b and c be centers. Then d(x,c) >= max\{0,
> d(x,b)-d(b,c)};
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}
> The application of Lemma 2 is a little complex: It need to cache/update the
> distance/lower bounds to previous centers, and thus can be only applied in
> training, not usable in prediction.
> So this ticket is mainly for Lemma 1. Its idea is quite simple, if point x is
> close to center b enough (less than a pre-computed radius), then we can say
> point x belong to center c without computing the distances between x and
> other centers. It can be used in both training and predction.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-31007) KMeans optimization based on triangle-inequality
[
https://issues.apache.org/jira/browse/SPARK-31007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17489235#comment-17489235
]
Sean R. Owen commented on SPARK-31007:
--
Hm, can you use a simpler "unpacked" representation for those statistics? it's
not the memory per se, but the size of a single array. Instead an array of
arrays?
> KMeans optimization based on triangle-inequality
>
>
> Key: SPARK-31007
> URL: https://issues.apache.org/jira/browse/SPARK-31007
> Project: Spark
> Issue Type: Improvement
> Components: ML
>Affects Versions: 3.1.0
>Reporter: zhengruifeng
>Assignee: zhengruifeng
>Priority: Major
> Fix For: 3.1.0
>
> Attachments: ICML03-022.pdf
>
>
> In current impl, following Lemma is used in KMeans:
> 0, Let x be a point, let b be a center and o be the origin, then d(x,c) >=
> |(d(x,o) - d(c,o))| = |norm(x)-norm(c)|
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}
> According to [Using the Triangle Inequality to Accelerate
> K-Means|[https://www.aaai.org/Papers/ICML/2003/ICML03-022.pdf]], we can go
> futher, and there are another two Lemmas can be used:
> 1, Let x be a point, and let b and c be centers. If d(b,c)>=2d(x,b) then
> d(x,c) >= d(x,b);
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}.
> However, luckily for CosineDistance we can get a variant in the space of
> radian/angle.
> 2, Let x be a point, and let b and c be centers. Then d(x,c) >= max\{0,
> d(x,b)-d(b,c)};
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}
> The application of Lemma 2 is a little complex: It need to cache/update the
> distance/lower bounds to previous centers, and thus can be only applied in
> training, not usable in prediction.
> So this ticket is mainly for Lemma 1. Its idea is quite simple, if point x is
> close to center b enough (less than a pre-computed radius), then we can say
> point x belong to center c without computing the distances between x and
> other centers. It can be used in both training and predction.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-31007) KMeans optimization based on triangle-inequality
[
https://issues.apache.org/jira/browse/SPARK-31007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17489230#comment-17489230
]
zhengruifeng commented on SPARK-31007:
--
[~srowen] This optimization needs an array of size
val packedValues = Array.ofDim[Double](k * (k + 1) / 2)
can may cause failure when k is large.
https://issues.apache.org/jira/browse/SPARK-36553 k=5
should we:
1, revert this optimization
2, or enable it only for small k? (but the impl will be more complex)
> KMeans optimization based on triangle-inequality
>
>
> Key: SPARK-31007
> URL: https://issues.apache.org/jira/browse/SPARK-31007
> Project: Spark
> Issue Type: Improvement
> Components: ML
>Affects Versions: 3.1.0
>Reporter: zhengruifeng
>Assignee: zhengruifeng
>Priority: Major
> Fix For: 3.1.0
>
> Attachments: ICML03-022.pdf
>
>
> In current impl, following Lemma is used in KMeans:
> 0, Let x be a point, let b be a center and o be the origin, then d(x,c) >=
> |(d(x,o) - d(c,o))| = |norm(x)-norm(c)|
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}
> According to [Using the Triangle Inequality to Accelerate
> K-Means|[https://www.aaai.org/Papers/ICML/2003/ICML03-022.pdf]], we can go
> futher, and there are another two Lemmas can be used:
> 1, Let x be a point, and let b and c be centers. If d(b,c)>=2d(x,b) then
> d(x,c) >= d(x,b);
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}.
> However, luckily for CosineDistance we can get a variant in the space of
> radian/angle.
> 2, Let x be a point, and let b and c be centers. Then d(x,c) >= max\{0,
> d(x,b)-d(b,c)};
> this can be applied in {{EuclideanDistance}}, but not in {{CosineDistance}}
> The application of Lemma 2 is a little complex: It need to cache/update the
> distance/lower bounds to previous centers, and thus can be only applied in
> training, not usable in prediction.
> So this ticket is mainly for Lemma 1. Its idea is quite simple, if point x is
> close to center b enough (less than a pre-computed radius), then we can say
> point x belong to center c without computing the distances between x and
> other centers. It can be used in both training and predction.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
