[jira] [Commented] (MADLIB-946) Null handling in kmeans

["Orhan Kislal (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MADLIB-946?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15109311#comment-15109311 ] 

Orhan Kislal commented on MADLIB-946:
-------------------------------------

I think it is possible to fix this with 2 changes. Update the closest column to return null instead of throwing an exception and update the kmeans to handle null as a valid centroid id. To the best of my knowledge, the exception in closest column isn't used for anything specific. 

> Null handling in kmeans
> -----------------------
>
>                 Key: MADLIB-946
>                 URL: https://issues.apache.org/jira/browse/MADLIB-946
>             Project: Apache MADlib
>          Issue Type: Bug
>          Components: Module: k-Means Clustering
>            Reporter: Frank McQuillan
>            Priority: Minor
>             Fix For: v1.9
>
>
> Opened for Yaw Tan:
> I'm using the Madlib k-means CLOSEST COLUMN function to determine the Cluster ID for the data point
> The error says
> Function:madlib.closest_column(double precision, double precision, regproc, text): Error converting an array w/ NULL values to dense format
> The fix is to ignore  NULL values in the array.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)