Bug in Optimize Parameter operator?
hi,
I was using k-nn inside optimize parameter (inside X-Validation) operator... and I was using euklidean and camberra distance for opt parameter settings...
but I noticed, that when using name "k-NN (2)" inside opt. parameter, it does not change the numerical_measure type... when I use a standard name "k-NN", it does change the parameters during running... seems to be a minor bug caused by the naming of the parameter as I experienced it?
Best Answers
-
Fred12 MemberPosts:344
Unicorn
ok here is a quick sample process where I used both k-NN:
in the one log the 2 distances show up, but not in the other:
<参数key="repository_entry" value="//Samples/data/Iris"/>
<参数key="number_of_validations" value="4"/>
<参数key="sampling_type" value="stratified sampling"/>
<参数key="gamma" value="0.007170375144641631"/>
<参数key="C" value="227500.0"/>
<连接from_port = "训练" to_op =“支持向量机”to_port ="training set"/>
< portSpacing端口= " sink_model”间隔= " 0 " / >
<参数key="kappa" value="true"/>
<参数key="weighted_mean_recall" value="true"/>
<参数key="weighted_mean_precision" value="true"/>
<参数key="correlation" value="true"/>
<参数key="squared_correlation" value="true"/>
<参数key="margin" value="true"/>
<参数key="repository_entry" value="//Samples/data/Iris"/>
<参数key="k-NN.k" value="[1.0;20;6;linear]"/>
<参数key="k-NN.numerical_measure" value="EuclideanDistance,CamberraDistance"/>
<参数key="k" value="20"/>
<参数key="numerical_measure" value="CamberraDistance"/>
< portSpacing端口= " sink_model”间隔= " 0 " / >
<参数key="kappa" value="true"/>
<参数key="weighted_mean_recall" value="true"/>
<参数key="weighted_mean_precision" value="true"/>
<参数key="correlation" value="true"/>
<参数key="squared_correlation" value="true"/>
<参数key="margin" value="true"/>
<参数key="k" value="operator.k-NN.parameter.k"/>
<参数key="nummeasure" value="operator.k-NN.parameter.numerical_measure"/>
<参数key="k-NN (2).k" value="[1.0;20;6;linear]"/>
<参数key="k-NN (2).numerical_measure" value="EuclideanDistance,CamberraDistance"/>
<参数key="k" value="20"/>
<参数key="numerical_measure" value="CamberraDistance"/>
< portSpacing端口= " sink_model”间隔= " 0 " / >
<参数key="kappa" value="true"/>
<参数key="weighted_mean_recall" value="true"/>
<参数key="weighted_mean_precision" value="true"/>
<参数key="correlation" value="true"/>
<参数key="squared_correlation" value="true"/>
<参数key="margin" value="true"/>
<参数key="k" value="operator.k-NN.parameter.k"/>
<参数key="nummeasure" value="operator.k-NN.parameter.numerical_measure"/>1 -
Thomas_Ott RapidMiner Certified Analyst, RapidMiner Certified Expert, MemberPosts:1761年
Unicorn
Thanks@Fred12for the process. I've tested it with a lower compatability too and the problem is the same. I have pinged the developers.
0
Answers
Hi,
Thanks for pointing this out. Can you do us a quick favor and post the XML of a process using a data set from the Sample repository? This would allow our engineers to replicate and chase this bug.
Thanks,
Ingo