+1Hi guys,
I just realized that the K-Means algorithm can support more distance methods of the algorithm than the default (euclidean), but the documentation doesn't say which ones support. I am not an algorithm expert, but I know that in other implementations you can use Manhattan and Cosine implementations (and I think there are more).
Which distance method options of the algorithm can be used in Vertica??
Thank you!
Regards,
David.
+2Hi,
I think only euclidean is supported now ...
dbadmin=> select version();
version
------------------------------------
Vertica Analytic Database v8.1.0-4
(1 row)
dbadmin=> SELECT KMEANS('myKmeansModel', 'iris1', '*', 5 USING PARAMETERS max_iterations=20, output_view='myKmeansView', key_columns='id', distance_method='jim', exclude_columns='Species');
ERROR 7506: Problem in kmeans.
Detail: Only 'euclidean' is supported for distance_method
+1+2Thanks David for pointing this out. You'll see in the latest doc that there is no reference to the distance method for this exact reason: http://docdev.verticacorp.com/doc/trunk/HTML/index.htm#Authoring/SQLReferenceManual/Functions/MachineLearning/KMEANS.htm?TocPath=SQL%20Reference%20Manual|SQL%20Functions|Machine%20Learning%20Functions|_____9
+1Oh, now I see. It is good to remove it from doc to not get confused. Thanks for the info!!
Anyway, I got some feedback from some Data Scientists that it would be good to have the option to change the distance method.
Already have an account? Login
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.
77 4th Avenue
Waltham, MA 02451 USA
