Hi Mark,
Yes, I am doing exactly the same thing you descibed, In the Experimenter, added test.rff as the data set and choose SMOReg with precomputed kernel as algorithm and gave kernel.txt as kernel matrix.
My small test.arff and kernel.txt is below,
But when I run with cross validation (2x1), the error is:
12:21:51: Class attribute is not nominal!
It seems as if it does not accept numeric values as target?
What do you suggest?
Thanks in advance,
test.arff:
@relation test
@attribute index {0,1,2}
@attribute target numeric
@data
0,1
1,2
2,3
kernel.txt:
3 3
1 0.5 0.2
0.5 1 0.5
0.2 0.5 1
Mark Hall-9 wrote:
On 15/03/12 3:20 AM, ahmeto wrote:
>
> Hi Mark,
> First of all,
> Thank you very much for the Weka version update, now I am able to use SMOReg
> with numeric attributes on Weka Explorer.
>
> But unfortunately, SMOReg does not work with numeric attributes in Weka
> Experimenter.
> What do you suggest for this problem? I believe Experimenter is using the
> same jars with Explorer, is not it?
>
> I need to use Experimenter since I want to use t-test and the values for
> each fold!
Hmm. Your small test example seemed to work ok for me in Experimenter.
What error are you seeing? I assume that you selected "Regression" in
the Setup panel of the Experimenter, and configured SMOreg with the
precomputed kernel matrix kernel + kernel file?
Cheers,
Mark.
_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@list.scms.waikato.ac.nz
List info and subscription status:
https://list.scms.waikato.ac.nz/mailman/listinfo/wekalistList etiquette:
http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
