Examples of org.apache.mahout.clustering.classify.ClusterClassifier

Package org.apache.mahout.clustering.classify

Examples of org.apache.mahout.clustering.classify.ClusterClassifier

org.apache.mahout.clustering.classify.ClusterClassifier
This classifier works with any ClusteringPolicy and its associated Clusters. It is initialized with a policy and a list of compatible clusters and thereafter it can classify any new Vector into one or more of the clusters based upon the pdf() function which each cluster supports. In addition, it is an OnlineLearner and can be trained. Training amounts to asking the actual model to observe the vector and closing the classifier causes all the models to computeParameters. Because a ClusterClassifier implements Writable, it can be written-to and read-from a sequence file as a single entity. For sequential and mapreduce clustering in conjunction with a ClusterIterator; however, it utilizes an exploded file format. In this format, the iterator writes the policy to a single POLICY_FILE_NAME file in the clustersOut directory and the models are written to one or more part-n files so that multiple reducers may employed to produce them.

    Configuration conf = new Configuration();
    FileSystem fs = FileSystem.get(pointsPath.toUri(), conf);
    List<VectorWritable> points = TestKmeansClustering.getPointsWritable(TestKmeansClustering.REFERENCE);
    ClusteringTestUtils.writePointsToFile(points, new Path(pointsPath, "file1"), fs, conf);
    Path path = new Path(priorPath, "priorClassifier");
    ClusterClassifier prior = newKlusterClassifier();
    prior.writeToSeqFiles(path);
    ClusteringPolicy policy = new KMeansClusteringPolicy();
    ClusterClassifier.writePolicy(policy, path);
    assertEquals(3, prior.getModels().size());
    System.out.println("Prior");
    for (Cluster cluster : prior.getModels()) {
      System.out.println(cluster.asFormatString(null));
    }
    new ClusterIterator().iterateMR(conf, pointsPath, path, outPath, 5);
    
    for (int i = 1; i <= 4; i++) {
      System.out.println("Classifier-" + i);
      ClusterClassifier posterior = new ClusterClassifier();
      String name = i == 4 ? "clusters-4-final" : "clusters-" + i;
      posterior.readFromSeqFiles(conf, new Path(outPath, name));
      assertEquals(3, posterior.getModels().size());
      for (Cluster cluster : posterior.getModels()) {
        System.out.println(cluster.asFormatString(null));
      }     
    }
  }

View Full Code Here

    }
  }
  
  @Test
  public void testCosineKlusterClassification() {
    ClusterClassifier classifier = newCosineKlusterClassifier();
    Vector pdf = classifier.classify(new DenseVector(2));
    assertEquals("[0,0]", "[0.333, 0.333, 0.333]", AbstractCluster.formatVector(pdf, null));
    pdf = classifier.classify(new DenseVector(2).assign(2));
    assertEquals("[2,2]", "[0.545, 0.273, 0.182]", AbstractCluster.formatVector(pdf, null));
  }

View Full Code Here

      runSequentialDirichletClusterer(input, output, modelDist, numClusters, numIterations, alpha0);
    } else {
      runSequentialDirichletClassifier(input, output, modelDist, numClusters, numIterations, alpha0);
    }
    for (int i = 1; i <= numIterations; i++) {
      ClusterClassifier posterior = new ClusterClassifier();
      String name = i == numIterations ? "clusters-" + i + "-final" : "clusters-" + i;
      posterior.readFromSeqFiles(new Configuration(), new Path(output, name));
      List<Cluster> clusters = Lists.newArrayList();
      for (Cluster cluster : posterior.getModels()) {
        if (isSignificant(cluster)) {
          clusters.add(cluster);
        }
      }
      CLUSTERS.add(clusters);

View Full Code Here

      throws IOException {
    List<Cluster> models = Lists.newArrayList();
    for (Model<VectorWritable> cluster : modelDist.sampleFromPrior(numClusters)) {
      models.add((Cluster) cluster);
    }
    ClusterClassifier prior = new ClusterClassifier(models, new DirichletClusteringPolicy(numClusters, alpha0));
    Path priorPath = new Path(output, Cluster.INITIAL_CLUSTERS_DIR);
    prior.writeToSeqFiles(priorPath);
    Configuration conf = new Configuration();
    new ClusterIterator().iterateSeq(conf, input, priorPath, output, numIterations);
  }

View Full Code Here

  }
  
  private List<List<Cluster>> getClusters(Path output, int numIterations) throws IOException {
    List<List<Cluster>> result = new ArrayList<List<Cluster>>();
    for (int i = 1; i <= numIterations; i++) {
      ClusterClassifier posterior = new ClusterClassifier();
      String name = i == numIterations ? "clusters-" + i + "-final" : "clusters-" + i;
      posterior.readFromSeqFiles(conf, new Path(output, name));
      List<Cluster> clusters = Lists.newArrayList();
      for (Cluster cluster : posterior.getModels()) {
        clusters.add(cluster);
      }
      result.add(clusters);
    }
    return result;

View Full Code Here

    List<Cluster> initialClusters = Lists.newArrayList();
    int id = 0;
    for (Vector point : points) {
      initialClusters.add(new org.apache.mahout.clustering.kmeans.Kluster(point, id++, measure));
    }
    ClusterClassifier prior = new ClusterClassifier(initialClusters, new KMeansClusteringPolicy(convergenceDelta));
    Path priorPath = new Path(output, Cluster.INITIAL_CLUSTERS_DIR);
    prior.writeToSeqFiles(priorPath);
    
    int maxIter = 10;
    new ClusterIterator().iterateSeq(conf, samples, priorPath, output, maxIter);
    loadClustersWritable(output);
  }

View Full Code Here

    List<Cluster> models = Lists.newArrayList();
    DistanceMeasure measure = new ManhattanDistanceMeasure();
    models.add(new DistanceMeasureCluster(new DenseVector(2).assign(1), 0, measure));
    models.add(new DistanceMeasureCluster(new DenseVector(2), 1, measure));
    models.add(new DistanceMeasureCluster(new DenseVector(2).assign(-1), 2, measure));
    return new ClusterClassifier(models, new KMeansClusteringPolicy());
  }

View Full Code Here

    List<Cluster> models = Lists.newArrayList();
    DistanceMeasure measure = new ManhattanDistanceMeasure();
    models.add(new org.apache.mahout.clustering.kmeans.Kluster(new DenseVector(2).assign(1), 0, measure));
    models.add(new org.apache.mahout.clustering.kmeans.Kluster(new DenseVector(2), 1, measure));
    models.add(new org.apache.mahout.clustering.kmeans.Kluster(new DenseVector(2).assign(-1), 2, measure));
    return new ClusterClassifier(models, new KMeansClusteringPolicy());
  }

View Full Code Here

    List<Cluster> models = Lists.newArrayList();
    DistanceMeasure measure = new CosineDistanceMeasure();
    models.add(new org.apache.mahout.clustering.kmeans.Kluster(new DenseVector(2).assign(1), 0, measure));
    models.add(new org.apache.mahout.clustering.kmeans.Kluster(new DenseVector(2), 1, measure));
    models.add(new org.apache.mahout.clustering.kmeans.Kluster(new DenseVector(2).assign(-1), 2, measure));
    return new ClusterClassifier(models, new KMeansClusteringPolicy());
  }

View Full Code Here

    List<Cluster> models = Lists.newArrayList();
    DistanceMeasure measure = new ManhattanDistanceMeasure();
    models.add(new SoftCluster(new DenseVector(2).assign(1), 0, measure));
    models.add(new SoftCluster(new DenseVector(2), 1, measure));
    models.add(new SoftCluster(new DenseVector(2).assign(-1), 2, measure));
    return new ClusterClassifier(models, new FuzzyKMeansClusteringPolicy());
  }

View Full Code Here

0 1 2 3 4 5 6 7 8 9

TOP

Related Classes of org.apache.mahout.clustering.classify.ClusterClassifier

org.apache.hadoop.conf.Configuration

org.apache.hadoop.fs.FileSystem

org.apache.hadoop.fs.Path

org.apache.hadoop.io.IntWritable

org.apache.hadoop.io.Text

org.apache.hadoop.io.Writable

org.apache.mahout.clustering.Cluster

org.apache.mahout.clustering.dirichlet.DirichletDriver

org.apache.mahout.clustering.dirichlet.TestL1ModelClustering

org.apache.mahout.clustering.dirichlet.TestMapReduce

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.