[IGNITE-9239] [ML] KMeansTrainer crashed if amount of possible clusters more than amount of partitions in dataset

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.7
• Component/s: ml
• Ignite Flags: Docs Required
• Assignee: Alexey Zinoviev
• Reporter: Alexey Zinoviev
• Created: 09/Aug/18 09:19
• Updated: 14/Jan/20 10:13
• Resolved: 16/Aug/18 12:56

Description

How to reproduce?

Set the K parameter in KMeans Trainer to 100, and run KMeansClusterization Example

\

StackTrace is

Exception in thread "KMeansClusterizationExample-#44" java.lang.RuntimeException: java.lang.IllegalArgumentException: bound must be positive
at org.apache.ignite.ml.clustering.kmeans.KMeansTrainer.fit(KMeansTrainer.java:112)
at org.apache.ignite.ml.clustering.kmeans.KMeansTrainer.fit(KMeansTrainer.java:46)
at org.apache.ignite.ml.trainers.DatasetTrainer.fit(DatasetTrainer.java:68)
at org.apache.ignite.examples.ml.clustering.KMeansClusterizationExample.lambda$main$0(KMeansClusterizationExample.java:60)
at java.lang.Thread.run(Thread.java:745)
Caused by: java.lang.IllegalArgumentException: bound must be positive
at java.util.Random.nextInt(Random.java:388)
at org.apache.ignite.ml.clustering.kmeans.KMeansTrainer.initClusterCentersRandomly(KMeansTrainer.java:193)
at org.apache.ignite.ml.clustering.kmeans.KMeansTrainer.fit(KMeansTrainer.java:86)
... 4 more

 

 

The possible solution :

correct the mechanism of rndPnts computation in the row 180-190 in KMeansTrainer

Issue Links

• Is contained by: IGNITE-9261 [ML] Add ANN algorithm based on ACD concept