Autoweka from matlab

[dolevr]

I'm trying to integrate autoweka with my code. I'm working from matlab environment and as it seems I get the following error when instantiating AutoWEKAClassifier.

Java exception occurred:
java.lang.IndexOutOfBoundsException: Index: 0, Size: 0
	at java.util.ArrayList.rangeCheck(Unknown Source)
	at java.util.ArrayList.get(Unknown Source)
	at autoweka.ApplicabilityTester.getApplicableClassifiers(ApplicabilityTester.java:62)
	at autoweka.ExperimentConstructor.loadClassifiers(ExperimentConstructor.java:336)
	at autoweka.ExperimentConstructor.run(ExperimentConstructor.java:210)
	at autoweka.ExperimentConstructor.buildSingle(ExperimentConstructor.java:158)
	at weka.classifiers.meta.AutoWEKAClassifier.buildClassifier(AutoWEKAClassifier.java:334)

This seem to an issue with instantiating the different weka classifiers even though they are found in the config:

weka.classifiers.bayes.BayesNet.params
weka.classifiers.bayes.NaiveBayes.params
weka.classifiers.bayes.NaiveBayesMultinomial.params
...

Any assistant will be appreciated ....

[larskotthoff]

How exactly are you using Auto-WEKA? What makes you think it's an issue with instantiating the classifiers?

[dolevr]

My mistake I confused it with an earlier problem I had. The issue arises when I try to perform training.

I had some wraper code (Java) for weka classifiers which I used for training/testing classifiers. Next I took the autoweka code and replaced the classifier instantiation with: clf = (Classifier) new AutoWEKAClassifier(); . When training, it causes the exception and stack trace above.

[larskotthoff]

Could you post a complete example that allows to reproduce the problem please?

[dolevr]

Sure:
Java:

public class MyWekaClassifier {
	private Classifier clf;
	public MyWekaClassifier(Config anomalyTagsConfig) throws Error {
		if (anomalyTagsConfig.useAutoWeka) {
			clf = (Classifier) new AutoWEKAClassifier();
		} else {
			initWekaClassifier(anomalyTagsConfig);
		}
		this.anomalyTagsConfig = anomalyTagsConfig;
	};

	public void classifierTrain(double[] inY, double[] inX, int numSamples, int numFeatures) throws Exception {
		Instances trainingData = data2Instances(inY, inX, numSamples, numFeatures);
		clf.buildClassifier(trainingData);
	}
}

The java class compiles to jar and stored in matlab working directory

Matlab code (which calls java):

javaaddpath('.\myJar.jar')
configOBJ = javaObject('com.myproject.Config');
configOBJ.useAutoWeka = config.useAutoWeka;

classifier = javaObject('com.myproject.MyWekaClassifier', configOBJ);
classifier.classifierTrain(yArray, reshape(trainData, [], 1), size(trainData, 2), size(trainData, 1));

I should also mention that I put all the config file in the same directory.

[larskotthoff]

I can't reproduce the problem. Here's a minimal class that does what you're doing. Running this code works fine for me:

import weka.classifiers.meta.AutoWEKAClassifier;
import weka.classifiers.Classifier;

public class Test {
    private Classifier clf;
	public Test() throws Error {
        clf = (Classifier) new AutoWEKAClassifier();
	};

    public static void main(String[] args) {
        Test test = new Test();
    }
}

[larskotthoff]

Closing as I can't reproduce the problem. Please reopen if you have a specific example that allows me to do so.