judy- | 1 Apr 02:55 2009

Re: Question about create an artificial data set


Here's my classifier setup:
weka.classifiers.bayes.BayesNet -D -Q
weka.classifiers.bayes.net.search.local.TAN -- -S ENTROPY -E
weka.classifiers.bayes.net.estimate.SimpleEstimator -- -A 0.5

You know TAN is based on conditional mutual information (CMI).In fact,in
weka I don't find CMI for TAN.To get the CMI scoring metric, I select
"Entropy" as scoreType for the search algorithm.Then the problem
aforementioned come up.

Thank you for your reply.

Best regards,
Judy

Peter Reutemann-3 wrote:
> 
>> I read Bayesian Network Classifiers in Weka for Version3-5-6.In the
>> explorer,I learned a network structure on the dataset
>> "weather.nominal.arff"
>> using TAN and saved the Bayes net in BIF format.The file is
>> weather.xml.Then
>> I want to create an artificial dataset using the Bayesnets ont the file.
>> In the command line I typed:
>> java weka.classifiers.bayes.net.BayesNetGenerator -F weather.xml -M 1000
>> And get this error:
>> java.lang.Exception: There appears to be a cycle in the graph
> 
> I just tried a similar scenario (not sure about your exact classifier
(Continue reading)

Peter Reutemann | 1 Apr 03:00 2009
Picon

Re: Question about create an artificial data set

> Here's my classifier setup:
> weka.classifiers.bayes.BayesNet -D -Q
> weka.classifiers.bayes.net.search.local.TAN -- -S ENTROPY -E
> weka.classifiers.bayes.net.estimate.SimpleEstimator -- -A 0.5
>
> You know TAN is based on conditional mutual information (CMI).

As a matter of fact, I don't. I don't have a bayesian background.

> In fact,in
> weka I don't find CMI for TAN.To get the CMI scoring metric, I select
> "Entropy" as scoreType for the search algorithm.Then the problem
> aforementioned come up.

Sorry, can't help there.

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
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

judy- | 1 Apr 03:07 2009

Re: Question about create an artificial data set


Thank you anyway.

Peter Reutemann-3 wrote:
> 
>> Here's my classifier setup:
>> weka.classifiers.bayes.BayesNet -D -Q
>> weka.classifiers.bayes.net.search.local.TAN -- -S ENTROPY -E
>> weka.classifiers.bayes.net.estimate.SimpleEstimator -- -A 0.5
>>
>> You know TAN is based on conditional mutual information (CMI).
> 
> As a matter of fact, I don't. I don't have a bayesian background.
> 
>> In fact,in
>> weka I don't find CMI for TAN.To get the CMI scoring metric, I select
>> "Entropy" as scoreType for the search algorithm.Then the problem
>> aforementioned come up.
> 
> Sorry, can't help there.
> 
> Cheers, Peter
> -- 
> Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
> http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174
> 
> _______________________________________________
> Wekalist mailing list
> Send posts to: Wekalist <at> list.scms.waikato.ac.nz
> List info and subscription status:
(Continue reading)

Wang, Jing | 1 Apr 03:20 2009
Picon

Question about the command line

Hi,

I am writing my codes using some of the weka classes. I call my class from the command line. Suppose the
command includes:

    -W weka.classifiers.functions.SMO -K weka.classifiers.functions.supportVector.PolyKernel

And I use the following codes to read the above command from the argument:

    String[] tmpOptions;
    String classname;
    tmpOptions     = Utils.splitOptions(Utils.getOption("W", args));
    classname      = tmpOptions[0];
    tmpOptions[0]  = "";
    Classifier cls = (Classifier) Utils.forName(Classifier.class, classname, tmpOptions);

So that it will get SMO as the classifier, and the parameters are shown in the tmpOptions, which here would be
"-K weka.classifiers.functions.supportVector.PolyKernel".

However, if I want to add a parameter to PolyKernel, such as change the exponent from the default value 1.0 to
2.0, then the part of the command line would be like:

    -W weka.classifiers.functions.SMO -K weka.classifiers.functions.supportVector.PolyKernel -- -E 2.0

But using the codes above to analyze the command wouldn't work, because "--" cannot be recognized. My
question is, could you possibly provide me with the updated codes which can read the command which
includes a parameter adding to the sub-classifier. I thought about it for a while, but still could not
figure out how to modify the codes.

Thanks,
(Continue reading)

Peter Reutemann | 1 Apr 03:23 2009
Picon

Re: Question about the command line

> I am writing my codes using some of the weka classes. I call my class from the command line. Suppose the
command includes:
>
>    -W weka.classifiers.functions.SMO -K weka.classifiers.functions.supportVector.PolyKernel
>
> And I use the following codes to read the above command from the argument:
>
>    String[] tmpOptions;
>    String classname;
>    tmpOptions     = Utils.splitOptions(Utils.getOption("W", args));
>    classname      = tmpOptions[0];
>    tmpOptions[0]  = "";
>    Classifier cls = (Classifier) Utils.forName(Classifier.class, classname, tmpOptions);
>
> So that it will get SMO as the classifier, and the parameters are shown in the tmpOptions, which here would
be "-K weka.classifiers.functions.supportVector.PolyKernel".
>
> However, if I want to add a parameter to PolyKernel, such as change the exponent from the default value 1.0
to 2.0, then the part of the command line would be like:
>
>    -W weka.classifiers.functions.SMO -K weka.classifiers.functions.supportVector.PolyKernel
-- -E 2.0
>
> But using the codes above to analyze the command wouldn't work, because "--" cannot be recognized. My
question is, could you possibly provide me with the updated codes which can read the command which
includes a parameter adding to the sub-classifier. I thought about it for a while, but still could not
figure out how to modify the codes.

You have to use the Utils.partitionOptions(...) method to obtain the
options after the "--". See the setOptions(String[]) method of the
(Continue reading)

Wang, Jing | 1 Apr 05:05 2009
Picon

RE: Question about the command line

Thank you. I've looked at the setOptions method in the SingleClassifierEnhancer class, but still cannot
think out a solution. 

In my case, I'll still use Utils.forName method, right? But what would I put in the argument "tmpOptions"?
It's an array of strings, a way I can think of is to put both "-K
weka.classifiers.functions.supportVector.PolyKernel" and "-E 2.0" into tmpOptions. But in this
way, when the Utils.forName method is performed, it will seem both options as belonging to the base
classifier SMO. That's not right. I want the option "-E 2.0" set under PolyKernel, not SMO. 

So how could I give the option containing PolyKernel which has already been set with the option "-E 2.0" to
SMO? Could you give me more details? 

Thanks again,
Jing

> I am writing my codes using some of the weka classes. I call my class from the command line. Suppose the
command includes:
>
>    -W weka.classifiers.functions.SMO -K weka.classifiers.functions.supportVector.PolyKernel
>
> And I use the following codes to read the above command from the argument:
>
>    String[] tmpOptions;
>    String classname;
>    tmpOptions     = Utils.splitOptions(Utils.getOption("W", args));
>    classname      = tmpOptions[0];
>    tmpOptions[0]  = "";
>    Classifier cls = (Classifier) Utils.forName(Classifier.class, classname, tmpOptions);
>
> So that it will get SMO as the classifier, and the parameters are shown in the tmpOptions, which here would
(Continue reading)

Peter Reutemann | 1 Apr 05:16 2009
Picon

Re: Question about the command line

> Thank you. I've looked at the setOptions method in the SingleClassifierEnhancer class, but still cannot
think out a solution.
>
> In my case, I'll still use Utils.forName method, right? But what would I put in the argument "tmpOptions"?
It's an array of strings, a way I can think of is to put both "-K
weka.classifiers.functions.supportVector.PolyKernel" and "-E 2.0" into tmpOptions. But in this
way, when the Utils.forName method is performed, it will seem both options as belonging to the base
classifier SMO. That's not right. I want the option "-E 2.0" set under PolyKernel, not SMO.
>
> So how could I give the option containing PolyKernel which has already been set with the option "-E 2.0" to
SMO? Could you give me more details?

Traditionally, the -W option takes only 1 argument, which is a
classname, and all the other options have to follow after the "--".

Your commandline should look something like this:
some.funky.Classifier -W weka.classifiers.functions.SMO <other
options> --  -K "weka.classifiers.functions.supportVector.PolyKernel
-E 2"

The classifier "some.funky.Classifier" would be derived from
SingleClassifierEnhancer. And SingleClassifierEnhancer automatically
uses the options after the "--"  and supplies the classifier specified
with -W (in this case, this is SMO).

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

(Continue reading)

Ed | 1 Apr 11:24 2009
Picon

IICAI-09 Call for papers

IICAI-09 Call for papers

 

The 4th Indian International Conference on Artificial Intelligence (IICAI-09) will be held in Tumkur (near Bangalore), India during December 16-18 2009. The conference consists of paper presentations, special workshops, sessions, invited talks and local tours, etc.  We invite draft paper submissions. Please see the website: http://www.iiconference.org   for more details of the conference.

 

Sincerely

 

 

Ed

Publicity Committee

 


_______________________________________________
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
utumno | 1 Apr 11:57 2009
Picon

Re: I want to processed reuters-21578 data


Peter Reutemann-3 wrote:
> 
>> I know we can download Reuters-21578 text data from UCI, but I am looking
>> for Reuters-21578 text categorization data compiled in arff format or
>> some
>> other format.
> 
> Have you had a look at the Weka homepage, "Datasets" section? Lists a
> link to the Reuters dataset.
> 
> 

I think in the datasets section there is the original reuters-21578 in text
form.
If you are looking for a different form (like the one svmlib or SVMLight
accepts), please contact me.
I have this dataset (and some more like OHSUMED, Reuters-RCV1, WebKB, 20NG,
TREC-AP) in this form. 
--

-- 
View this message in context: http://www.nabble.com/I-want-to-processed-reuters-21578-data-tp22734052p22823252.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
utumno | 1 Apr 12:00 2009
Picon

CSVLoader memory requirements


Greetings,

I am running this simple code.

CSVLoader csvlTr = new CSVLoader();
csvlTr.setSource(new File("train.csv"));
Instances train = csvlTr.getDataSet();
System.err.println(train.toSummaryString());

train.csv is a tab separated dataset. It consists of 25,000 lines. The
number of attributes is 15,000. All attributes are numerical. The size of
this file is about 1.6GB. A simple calculation of "25,000 examples x 15,000 
attributes x 8 bytes per double value" would result that this dataset
requires about 3GB of memory.

The strange thing is that this code whould never ends, although I have 6 GB
of physical RAM and 12GB of swap space. After about an hour the system hangs
up having consumed all of its swap space. 

Thanks in advance,

Anestis Gk
--

-- 
View this message in context: http://www.nabble.com/CSVLoader-memory-requirements-tp22823302p22823302.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

Gmane