Examples of cc.mallet.types.Instance

cc.mallet.types.Instance
A machine learning "example" to be used in training, testing or performance of various machine learning algorithms.
An instance contains four generic fields of predefined name: "data", "target", "name", and "source". "Data" holds the data represented `by the instance, "target" is often a label associated with the instance, "name" is a short identifying name for the instance (such as a filename), and "source" is human-readable sourceinformation, (such as the original text).
Each field has no predefined type, and may change type as the instance is processed. For example, the data field may start off being a string that represents a file name and then be processed by a {@link cc.mallet.pipe.Pipe} into a CharSequencerepresenting the contents of the file, and eventually to a feature vector holding indices into an {@link cc.mallet.types.Alphabet} holding words found in the file.It is up to each pipe which fields in the Instance it modifies; the most common case is that the pipe modifies the data field.
Generally speaking, there are two modes of operation for Instances. (1) An instance gets created and passed through a Pipe, and the resulting data/target/name/source fields are used. This is generally done for training instances. (2) An instance gets created with raw values in its slots, then different users of the instance call newPipedCopy() with their respective different pipes. This might be done for test instances at "performance" time.
Rather than store an {@link cc.mallet.types.Alphabet} in the Instance,we obtain it through the Pipe instance variable, because the Pipe also indicates where the data came from and how to interpret the Alphabet.
Instances can be made immutable if locked. Although unlocked Instances are mutable, typically the only code that changes the values in the four slots is inside Pipes.
Note that constructing an instance with a pipe argument means "Construct the instance and then run it through the pipe". {@link cc.mallet.types.InstanceList} uses this methodwhen adding instances through a pipeInputIterator. @see Pipe @see Alphabet @see InstanceList @author Andrew McCallum mccallum@cs.umass.edu

      if (ilist == null)
        throw new IllegalStateException ("Frozen.  Cannot split.");
      InstanceList ilist0 = new InstanceList (ilist.getPipe());
      InstanceList ilist1 = new InstanceList (ilist.getPipe());
      for (int i = 0; i < ilist.size(); i++) {
        Instance instance = ilist.get(i);
        FeatureVector fv = (FeatureVector) instance.getData ();
        // xxx What test should this be?  What to do with negative values?
          // Whatever is decided here should also go in InfoGain.calcInfoGains()
        if (fv.value (featureIndex) != 0) {
          //System.out.println ("list1 add "+instance.getUri()+" weight="+ilist.getInstanceWeight(i));
          ilist1.add (instance, ilist.getInstanceWeight(i));

View Full Code Here

      MatrixOps.expNormalize(scores);
   
      entropy += Maths.getEntropy(scores);


      LabelVector lv = new LabelVector((LabelAlphabet)data.getTargetAlphabet(), scores);
      Instance instance = new Instance(data.get(ii).getData(),lv,null,null);
      dataLabeled.add(instance);
    }
    
    // train supervised
    MaxEntOptimizableByLabelDistribution opt = new  MaxEntOptimizableByLabelDistribution(dataLabeled,p);

View Full Code Here

    double[] results = new double[numLabels];
    for (int iter = 0; iter < m_maxIterations; iter++) {


      // loop through all instances
      for (int ii = 0; ii < trainingList.size(); ii++) {
        Instance inst = trainingList.get(ii);
        Labeling labeling = inst.getLabeling ();
        FeatureVector fv = (FeatureVector) inst.getData();
        int fvisize = fv.numLocations();
        int correctIndex = labeling.getBestIndex();
        Arrays.fill(results, 0);


        // compute dot(x, wi) for each class i

View Full Code Here


        if (testSequence.size() < 1) {
            return new Pair<Double, Sequence<?>>(-1.0, null);
        }


        Instance inst = testSequence.get(0);
        input = (Sequence<?>) inst.getData();


        output = crf.transduce(input);
        conf = crf_estimator.estimateConfidenceFor(inst, startTags, inTags);


        return new Pair<Double, Sequence<?>>(conf, output);

View Full Code Here

  private boolean[][] labelConnectionsIn (InstanceList trainingSet, String start)
  {
    int numLabels = outputAlphabet.size();
    boolean[][] connections = new boolean[numLabels][numLabels];
    for (int i = 0; i < trainingSet.size(); i++) {
      Instance instance = trainingSet.get(i);
      FeatureSequence output = (FeatureSequence) instance.getTarget();
      for (int j = 1; j < output.size(); j++) {
        int sourceIndex = outputAlphabet.lookupIndex (output.get(j-1));
        int destIndex = outputAlphabet.lookupIndex (output.get(j));
        assert (sourceIndex >= 0 && destIndex >= 0);
        connections[sourceIndex][destIndex] = true;

View Full Code Here

    for (int i = 0; i < parameters.weights.length; i++) 
      for (int j = parameters.weights[i].numLocations()-1; j >= 0; j--)
        weightsPresent[i].set (parameters.weights[i].indexAtLocation(j));
    // Put in the weights in the training set
    for (int i = 0; i < trainingData.size(); i++) {
      Instance instance = trainingData.get(i);
      FeatureVectorSequence input = (FeatureVectorSequence) instance.getData();
      FeatureSequence output = (FeatureSequence) instance.getTarget();
      // gsc: trainingData can have unlabeled instances as well
      if (output != null && output.size() > 0) {
        // Do it for the paths consistent with the labels...
        sumLatticeFactory.newSumLattice (this, input, output, new Transducer.Incrementor() {
          public void incrementTransition (Transducer.TransitionIterator ti, double count) {

View Full Code Here

      FeatureInducer klfi = (FeatureInducer)featureInducers.get(i);
      klfi.induceFeaturesFor (testing, false, false);
    }
    Sequence[] ret = new Sequence[testing.size()];
    for (int i = 0; i < testing.size(); i++) {
      Instance instance = testing.get(i);
      Sequence input = (Sequence) instance.getData();
      Sequence trueOutput = (Sequence) instance.getTarget();
      assert (input.size() == trueOutput.size());
      Sequence predOutput = new MaxLatticeDefault(this, input).bestOutputSequence();
      assert (predOutput.size() == trueOutput.size());
      ret[i] = predOutput;
    }

View Full Code Here

        new SGML2TokenSequence()
//        new SGML2TokenSequence (new CharSequenceLexer (Pattern.compile (".")), "O")
        });


      for (int i = 0; i < args.length; i++) {
        Instance carrier = p.instanceFrom(new Instance (new File(args[i]), null, null, null));
        TokenSequence data = (TokenSequence) carrier.getData();
        TokenSequence target = (TokenSequence) carrier.getTarget();
        logger.finer ("===");
        logger.info (args[i]);
        for (int j = 0; j < data.size(); j++)
          logger.info (target.get(j).getText()+" "+data.get(j).getText());
      }

View Full Code Here


    double[][] scores = new double[trainingList.size()][numLabels];
    
    // pass 1: calculate model distribution
    for (int ii = 0; ii < trainingList.size(); ii++) {
      Instance instance = trainingList.get(ii);
      double instanceWeight = trainingList.getInstanceWeight(instance);
      
      // skip if labeled
      if (instance.getTarget() != null) {
        continue;
      }
      
      FeatureVector fv = (FeatureVector) instance.getData();
      classifier.getClassificationScoresWithTemperature(instance, temperature, scores[ii]);
      
      for (int loc = 0; loc < fv.numLocations(); loc++) {
        int featureIndex = fv.indexAtLocation(loc);
        if (constraints.containsKey(featureIndex)) {
          int cIndex = mapping.get(featureIndex);            
          double val;
          if (!useValues) {
            val = 1.;
          }
          else {
            val = fv.valueAtLocation(loc);
          }
          featureCounts[cIndex] += val;
          for (int l = 0; l < numLabels; l++) {
            modelExpectations[cIndex][l] += scores[ii][l] * val * instanceWeight;
          }
        }
      }
      
      // special case of label regularization
      if (constraints.containsKey(defaultFeatureIndex)) {
        int cIndex = mapping.get(defaultFeatureIndex); 
        featureCounts[cIndex] += 1;
        for (int l = 0; l < numLabels; l++) {
          modelExpectations[cIndex][l] += scores[ii][l] * instanceWeight;
        }        
      }
    }
    
    double value = 0;
    for (int featureIndex : constraints.keySet()) {
      int cIndex = mapping.get(featureIndex);
      if (featureCounts[cIndex] > 0) {
        for (int label = 0; label < numLabels; label++) {
          double cProb = constraints.get(featureIndex)[label];
          // normalize by count
          modelExpectations[cIndex][label] /= featureCounts[cIndex];
          ratio[cIndex][label] =  cProb / modelExpectations[cIndex][label];
          // add to the cross entropy term
          value += scalingFactor * cProb * Math.log(modelExpectations[cIndex][label]);
          // add to the entropy term
          if (cProb > 0) {
            value -= scalingFactor * cProb * Math.log(cProb);
          }
        }
        assert(Maths.almostEquals(MatrixOps.sum(modelExpectations[cIndex]),1));
      }
    }


    // pass 2: determine per example gradient
    for (int ii = 0; ii < trainingList.size(); ii++) {
      Instance instance = trainingList.get(ii);
      
      // skip if labeled
      if (instance.getTarget() != null) {
        continue;
      }
      
      double instanceWeight = trainingList.getInstanceWeight(instance);
      FeatureVector fv = (FeatureVector) instance.getData();


      for (int loc = 0; loc < fv.numLocations() + 1; loc++) {
        int featureIndex;
        if (loc == fv.numLocations()) {
          featureIndex = defaultFeatureIndex;

View Full Code Here


    Iterator<Integer> keyIter = labeledFeatures.keySet().iterator();
    
    double[][] featureCounts = new double[labeledFeatures.size()][numLabels];
    for (int ii = 0; ii < trainingData.size(); ii++) {
      Instance instance = trainingData.get(ii);
      FeatureVector fv = (FeatureVector)instance.getData();
      Labeling labeling = trainingData.get(ii).getLabeling();
      double[] labelDist = new double[numLabels];
      
      if (labeling == null) {
        labelByVoting(labeledFeatures,instance,labelDist);

View Full Code Here

0 1 2 3 4 5 6 7 8 9

TOP

Related Classes of cc.mallet.types.Instance

cc.mallet.classify.BalancedWinnowTrainer

cc.mallet.classify.C45$Node

cc.mallet.classify.Classification

cc.mallet.classify.Classifier

cc.mallet.classify.DecisionTree$Node

cc.mallet.classify.evaluate.ConfusionMatrix

cc.mallet.classify.FeatureConstraintUtil

cc.mallet.classify.MaxEntOptimizableByGE

cc.mallet.classify.MaxEntOptimizableByLabelDistribution

cc.mallet.classify.MaxEntOptimizableByLabelLikelihood

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.