WEKA machine learning software discussion

Rizwan Mian | 6 Aug 2012 23:00

non-linear regression with SMOReg?

Hi Folks,

Are there any tutorials on using SMOReg for non-linear data with Weka?

I am also seeking input on how would user provide suitable values for parameters such as c; kernel parameters (exponent, gamma); regOptimizer (epsilon, eplisonparameter)? Examples would really help.

Finally, intuitive explaination for the role of these parameters would really help.

thanking in advance. :-)

--
Rizwan Mian
Doctoral Candidate
School of Computing (Sum Ergo Computo)
Queens University
http://research.cs.queensu.ca/~mian/index.htm

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

headers

Mark Hall | 10 Aug 2012 06:00

Re: non-linear regression with SMOReg?

On 7/08/12 9:00 AM, Rizwan Mian wrote:
> Hi Folks,
>
> Are there any tutorials on using SMOReg for non-linear data with Weka?
>
> I am also seeking input on how would user provide suitable values for
> parameters such as c; kernel parameters (exponent, gamma); regOptimizer
> (epsilon, eplisonparameter)? Examples would really help.
>
> Finally, intuitive explaination for the role of these parameters would
> really help.
>
> thanking in advance. 

Sections 5 and 6 of this article might be of use:

https://docs.google.com/viewer?a=v&q=cache:G9J0DanvCnQJ:pyml.sourceforge.net/doc/howto.pdf+&hl=en&gl=nz&pid=bl&srcid=ADGEEShH5FCWxX7bnhGaoXnb2auqLChg0ndpgsEzqpaAzhH-OwqIHZ9nwvpjAnW2hJ5QUegTC9f5CcD7DGtl1evlRKIBIFezaAsyPX11unR9WohJDqNhtPU-zqg1hDVP8R_cIacIRbmH&sig=AHIEtbT7mx9rcURikPnhr4sXQ19dQcn4qQ

Cheers,
Mark.

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

headers

vizvan | 13 Aug 2012 17:51

Re: non-linear regression with SMOReg?


> Sections 5 and 6 of this article might be of use: ...

Thanks Mark. Looks relevant. I'll take a detailed look.

My investigation on the parameters for the interested readers.
discussion on SVM parameters [40]
Optimizing parameters using CVParameterSelection or GridSearch [41]: example
of Optimizing SMO with RBFKernel (C and gamma) but works with 3.6.5 not with
3.7.5 due to an exception [42]

Rizwan 

[40] http://www.svms.org/parameters/
[41] http://weka.wikispaces.com/Optimizing+parameters
[42]
https://list.scms.waikato.ac.nz/pipermail/wekalist/2011-April/052033.html

--

-- 
View this message in context: http://old.nabble.com/non-linear-regression-with-SMOReg--tp34263333p34291585.html
Sent from the WEKA mailing list archive at Nabble.com.

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

headers

Mark Hall | 14 Aug 2012 00:14

Re: non-linear regression with SMOReg?

On 14/08/12 3:51 AM, vizvan wrote:
>
>> Sections 5 and 6 of this article might be of use: ...
>
> Thanks Mark. Looks relevant. I'll take a detailed look.
>
> My investigation on the parameters for the interested readers.
> discussion on SVM parameters [40]
> Optimizing parameters using CVParameterSelection or GridSearch [41]: example
> of Optimizing SMO with RBFKernel (C and gamma) but works with 3.6.5 not with
> 3.7.5 due to an exception [42]
>
> Rizwan 
>
> [40] http://www.svms.org/parameters/
> [41] http://weka.wikispaces.com/Optimizing+parameters
> [42]
> https://list.scms.waikato.ac.nz/pipermail/wekalist/2011-April/052033.html
>
>

Hi Rizwan,

Would you be able to provide a data set and settings for GridSearch that 
generate the IllegalStateException? We need a data set that demonstrates 
this so that we can debug.

Cheers,
Mark.

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

headers

vizvan | 14 Aug 2012 11:15

Re: non-linear regression with SMOReg?


> Would you be able to provide a data set and settings for GridSearch that 
> generate the IllegalStateException? We need a data set that demonstrates 
> this so that we can debug.

It has been a while. Nonetheless, I attach two data sets. Larger data set is
used for training and the smaller for evaluation. I performed a grid search
over C and gemma parameter, similar to [1]. The C parameter is varied from
-5 to 15, while gemma is varied from -15 to 3. Both C and gemma have base of
2, step of 1 and expression of "pow(BASE,I)" as suggested by [2]. The
evaluation metric is the correlation coefficient.

Please, keep me posted. 

[1] http://weka.wikispaces.com/Optimizing+parameters
[2] Akay, M.F., 2009. Support vector machines combined with feature
selection for breast cancer diagnosis. Expert Systems with Applications 36,
2, Part 2, 3240-3247. DOI= http://dx.doi.org/10.1016/j.eswa.2008.01.009.

http://old.nabble.com/file/p34295778/input%252Bthroughput_trade-order.csv
input%2Bthroughput_trade-order.csv 
http://old.nabble.com/file/p34295778/input%252Bthroughput.csv_trade-order.csv
input%2Bthroughput.csv_trade-order.csv 
--

-- 
View this message in context: http://old.nabble.com/non-linear-regression-with-SMOReg--tp34263333p34295778.html
Sent from the WEKA mailing list archive at Nabble.com.

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html