<?xml version="1.0" encoding="utf-8"?>
<feed xml:lang="en" xmlns="http://www.w3.org/2005/Atom"><title>Recent posts to Discussion</title><link href="https://sourceforge.net/p/meka/discussion/" rel="alternate"/><link href="https://sourceforge.net/p/meka/discussion/feed.atom" rel="self"/><id>https://sourceforge.net/p/meka/discussion/</id><updated>2019-04-03T15:57:59.527000Z</updated><subtitle>Recent posts to Discussion</subtitle><entry><title>Running  MEKA on GPU</title><link href="https://sourceforge.net/p/meka/discussion/help/thread/4f1a4c30f0/?limit=25#69fe" rel="alternate"/><published>2019-04-03T15:57:59.527000Z</published><updated>2019-04-03T15:57:59.527000Z</updated><author><name>Yogesh </name><uri>https://sourceforge.net/u/sayogi/</uri></author><id>https://sourceforge.net9e519f9c57141df98b0fa5a3fa55efdd3b5c6a35</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hello Everyone,&lt;br/&gt;
Is it possible to run MEKA  on a GPU?&lt;br/&gt;
For instance, is it possible to run the CLI for MEKA to somehow use the power of GPU instead of letting it run on CPU? As running MEKA on CPU sometime is very time consuming especially with huge amount of data and some particulat algorithms like kNN or Random forest. I know that Machine learninf takes time but still if there is any way to explore the power of GPU for MEKA it will be great.  Can that be done? &lt;br/&gt;
Any inputs will be higly appreciated.&lt;br/&gt;
thanks in advance&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Wrong Predictions</title><link href="https://sourceforge.net/p/meka/discussion/help/thread/1e04c128c3/?limit=25#6bc6" rel="alternate"/><published>2019-01-14T21:38:16.204000Z</published><updated>2019-01-14T21:38:16.204000Z</updated><author><name>mali</name><uri>https://sourceforge.net/u/malsh20/</uri></author><id>https://sourceforge.net4f06cefc5ac22107ff3b8d9951315d4fb71ed6b5</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hi&lt;br/&gt;
I need to do multi target regression with 7 targets, also my data is updateable then I used CRUpdateable classifier to have a updateable training data as stream.&lt;br/&gt;
CRUpdateable classifier = new CRUpdateable();&lt;br/&gt;
SGD sgd = new SGD(); // weka calssifier&lt;br/&gt;
classifier.setClassifier(sgd);&lt;br/&gt;
then used Evaluation Class to evaluate my model.&lt;br/&gt;
Result result = Evaluation.evaluateModel((MultiXClassifier) classifier , dataForEval , top, vop);&lt;/p&gt;
&lt;p&gt;but when I get the prediction from result object it gives me some strange numbers which I've found they are &lt;br/&gt;
prediction confidence not prediction.&lt;br/&gt;
they are so close together ....&lt;/p&gt;
&lt;p&gt;ArrayList&amp;lt;double[]&amp;gt; predictions = result.predictions;&amp;lt;/double[]&amp;gt;&lt;/p&gt;
&lt;p&gt;545.4913271745663  17.284110801102905  4.818136346271051   94.40754051723867   14.099466445155207  10.068466645891423  3.6345297343715037&lt;br/&gt;
545.5345216099157   17.262783404046566  4.79602257368674    94.3954113751965    14.077618154557285  10.047245881674723  3.612201987273436 &lt;br/&gt;
545.5252465592279   17.2782622973221    4.811673925908964   94.40777188205296   14.092659802876382  10.062264373542853  3.6277189844752513&lt;br/&gt;
545.5496634176129   17.27402964495623   4.807847602102359   94.40780265069291   14.088930011196364  10.058984737738813  3.623851295809089 &lt;br/&gt;
545.4993549697862   17.25950998975275   4.796344849166108   94.38732180865863   14.073982665402944  10.047129253599238  3.612860995527209 &lt;br/&gt;
545.3085390338642   17.232835812581413  4.774505754083781   94.33362697516067   14.049400939955069  10.019842826649256  3.5909241452022926&lt;/p&gt;
&lt;p&gt;But I need prediction not prediction confidence, how I can get the true predictions?&lt;br/&gt;
 Please advise,&lt;br/&gt;
 Thanks&lt;br/&gt;
 Mali&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Instance Class Index</title><link href="https://sourceforge.net/p/meka/discussion/help/thread/72c960eff6/?limit=25#0a53" rel="alternate"/><published>2018-10-20T07:02:17.785000Z</published><updated>2018-10-20T07:02:17.785000Z</updated><author><name>mali</name><uri>https://sourceforge.net/u/malsh20/</uri></author><id>https://sourceforge.net668a8e1683ad0f8d81ebda91b9efd0a5e8c2751d</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hi,&lt;/p&gt;
&lt;p&gt;I want to do MultiTargetReression by Meka and my code is as under:&lt;/p&gt;
&lt;div class="codehilite"&gt;&lt;pre&gt;&lt;span&gt;&lt;/span&gt;    try {
        ConverterUtils.DataSource dataSource = new ConverterUtils.DataSource(FILE_PATH);
        Instances preparedDataSet = dataSource.getDataSet();
        preparedDataSet = filterUnsupervisedAttributes(preparedDataSet);
        preparedDataSet.setClassIndex(8);

        Instances  trainingInstances = new Instances(dataSource.getStructure());
        trainingInstances = filterUnsupervisedAttributes(trainingInstances);
        trainingInstances.setClassIndex(8);

        CRUpdateable classifier = new CRUpdateable();
        RandomForest randomForest = new RandomForest();
        classifier.setClassifier(randomForest);

        int numInst = preparedDataSet.numInstances();
        for(int row = 0; row &amp;lt; numInst; row++) {
            Instance instance = preparedDataSet.instance(row);
            System.out.println("instance index : "+instance.classIndex());
            trainingInstances.add(instance);
            if (row != 0 &amp;amp;&amp;amp; row%100==0) {

                classifier.buildClassifier(trainingInstances);

            }

        }

    } catch (Exception e) {
        e.printStackTrace();
    }

    With this code I want to build classifier 100 by 100, to do thet, I create the new Instances object as trainingInstances and put 100 intances then call buildClassifier but it has some error as under:

    weka.core.WekaException: weka.classifiers.trees.RandomTree: Not enough training instances with class labels (required: 1, provided: 0)!
at weka.core.Capabilities.test(Capabilities.java:1286)
at weka.core.Capabilities.test(Capabilities.java:1138)
at weka.core.Capabilities.testWithFail(Capabilities.java:1468)
at weka.classifiers.meta.Bagging.buildClassifier(Bagging.java:696)
at meka.classifiers.multitarget.CR.buildClassifier(CR.java:81)

    You need to know if I am giving preparedDataSet to buildClassifier which is original dataset there is no errors. but I need to train the model 100 by 100.

   Appreciate if you can help me.

   Thanks
   Mali
&lt;/pre&gt;&lt;/div&gt;

&lt;/div&gt;</summary></entry><entry><title>Startup error run.sh</title><link href="https://sourceforge.net/p/meka/discussion/help/thread/06ff9c03/?limit=25#8b32" rel="alternate"/><published>2018-09-05T17:01:35.267000Z</published><updated>2018-09-05T17:01:35.267000Z</updated><author><name>Vlad C</name><uri>https://sourceforge.net/u/marocanu/</uri></author><id>https://sourceforge.net530648ac7a9e926f5f7be1265a1f00c1c609be61</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hello, I'm encountering the following error when starting the run.sh app:&lt;/p&gt;
&lt;p&gt;Vlads-MacBook:~ vlad$ /Users/vlad/Downloads/meka-release-1.9.0/run.sh ; exit;\
Exception in thread "main" java.lang.NoClassDefFoundError: meka/gui/guichooser/GUIChooser\
Caused by: java.lang.ClassNotFoundException: meka.gui.guichooser.GUIChooser\
    at java.net.URLClassLoader$1.run(URLClassLoader.java:202)\
    at java.security.AccessController.doPrivileged(Native Method)\
    at java.net.URLClassLoader.findClass(URLClassLoader.java:190)\
    at java.lang.ClassLoader.loadClass(ClassLoader.java:306)\
    at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:301)\
    at java.lang.ClassLoader.loadClass(ClassLoader.java:247)\
logout\
Saving session...\
...copying shared history...\
...saving history...truncating history files...\
...completed.\
Deleting expired sessions...40 completed.\
\
&lt;span&gt;[Process completed]&lt;/span&gt;\
}&lt;/p&gt;
&lt;p&gt;Does anyone have any idea what i'm doing wrong? Could it be that some class is missing from the downloaded meka package?&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Newbie help needed</title><link href="https://sourceforge.net/p/meka/discussion/help/thread/e0c45afa/?limit=25#1cef" rel="alternate"/><published>2018-08-08T02:40:29.146000Z</published><updated>2018-08-08T02:40:29.146000Z</updated><author><name>chanda</name><uri>https://sourceforge.net/u/chandabnd727/</uri></author><id>https://sourceforge.net70d1ea52755f6920b1b649ef44993ba6f7cd56cc</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;I am a newbie to MEKA and WEKA.&lt;/p&gt;
&lt;p&gt;I have files with data that are labelled with multiple labels.&lt;/p&gt;
&lt;p&gt;e.g. "This is a sentence" {Angry, Sad, Happy}&lt;/p&gt;
&lt;p&gt;I need to get precision and recall for KNN , SVM and a few others.&lt;/p&gt;
&lt;p&gt;Being a beginner to MEKA I have no idea how to get going on this.&lt;/p&gt;
&lt;p&gt;I have used WEKA and the CLI and I know about ARFF files.&lt;/p&gt;
&lt;p&gt;But I dont have any clue how to make MEKA give me figures&lt;br/&gt;
for multiLABEL classification.&lt;/p&gt;
&lt;p&gt;Thank you kindly.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Multi-label classification using MLkNN via Meka </title><link href="https://sourceforge.net/p/meka/discussion/help/thread/042e3717/?limit=25#c934" rel="alternate"/><published>2018-08-08T00:03:14.557000Z</published><updated>2018-08-08T00:03:14.557000Z</updated><author><name>chanda</name><uri>https://sourceforge.net/u/chandabnd727/</uri></author><id>https://sourceforge.net3c72e38b2c2a9117ab07cac971dbf40c346f9537</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;MEKA is built on WEKA, WEKA takes a -x parameter to specify the number of folds. Could that be what you are looking for?&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Multi-label classification using MLkNN via Meka </title><link href="https://sourceforge.net/p/meka/discussion/help/thread/042e3717/?limit=25#75aa" rel="alternate"/><published>2018-08-06T10:55:20.666000Z</published><updated>2018-08-06T10:55:20.666000Z</updated><author><name>Adnan Shah</name><uri>https://sourceforge.net/u/iamadnan/</uri></author><id>https://sourceforge.net73f65d2b6da483394fce0460ec984836e179742a</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hi,&lt;br/&gt;
I am new to data mining and have doubts on understanding what this excerpt from the book&lt;strong&gt; Multi-Label Classification Problem Analysis and Metrics&lt;/strong&gt; actually means.&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;The MLDs were partitioned following a 2×5 strategy. This means that there are two repetitions with 5 folds, and that for each run 80% (4/5) of instances are &amp;gt;used for training and 20%(1/5) for testing. Therefore, a total of 10 runs are made for  each MLD. Random sampling was used to select the instances in each fold. The full set of folds for the aforementioned five MLDs is available in the book repository&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;So what I get is a 10 train/test datasets in separate files, so....&lt;br/&gt;
Q. Has the author already manually partitioning the single dataset into 5 folds and now the partitioned datasets (train/test) need to be directly classified one-by-one?&lt;/p&gt;
&lt;p&gt;OR&lt;/p&gt;
&lt;p&gt;Is he referring to some automated partitioning and then batch classification via MEKA, but as far as I have searched in MEKA I couldn't find a functionality to partition a dataset in folds plus train/test splits in one single operation.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Exception</title><link href="https://sourceforge.net/p/meka/discussion/help/thread/a816f3e5/?limit=25#b62e" rel="alternate"/><published>2018-07-06T09:14:32.742000Z</published><updated>2018-07-06T09:14:32.742000Z</updated><author><name>sara</name><uri>https://sourceforge.net/u/ss2018/</uri></author><id>https://sourceforge.net73b096001d96847db54eec3e32a7e5596c2ae7b2</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Hi  everybody!&lt;br/&gt;
I'm new in meka and work with GUI.when i load  dataset in MULAN like CAL500 and choose classify and start it i have this error!&lt;br/&gt;
&lt;strong&gt;Evaluation Failed!java.lang.index Out Of Bounds Exception:index:0,Size:0&lt;/strong&gt;&lt;br/&gt;
what can i do?&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Adding multi-label classifier code written in java to MEKA</title><link href="https://sourceforge.net/p/meka/discussion/help/thread/e025c65c/?limit=25#3181" rel="alternate"/><published>2018-04-15T13:28:08.334000Z</published><updated>2018-04-15T13:28:08.334000Z</updated><author><name>Vaishali</name><uri>https://sourceforge.net/u/vstsf/</uri></author><id>https://sourceforge.net1866e666c7adf82e52db47989e5e6c1e98ba53c3</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Sir,&lt;br/&gt;
      I am trying to add multi-label classifier code written in java to MEKA. I have used separate train and test files. But now I want to perform 10 fold cross validation in MEKA. is it possible?&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>How to get the precision and recall of each label?</title><link href="https://sourceforge.net/p/meka/discussion/general/thread/41885d7d/?limit=25#4a73" rel="alternate"/><published>2017-11-16T10:16:33.790000Z</published><updated>2017-11-16T10:16:33.790000Z</updated><author><name>Jawad Shafi</name><uri>https://sourceforge.net/u/jawadsm/</uri></author><id>https://sourceforge.net811bed19f4b68b5be6a34758a0a946d30a4d90fb</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Thanks a lot, indeed. &lt;/p&gt;&lt;/div&gt;</summary></entry></feed>