Package org.apache.hama.ml.math

Examples of org.apache.hama.ml.math.DenseDoubleVector

183
184
185
186
187
188
189
      String[] split = s.split(" ");
      double[] dar = new double[split.length];
      for (int i = 0; i < split.length; i++) {
        dar[i] = Double.valueOf(split[i]);
      }
      return new DenseDoubleVector(dar);
    }
View Full Code Here
176
177
178
179
180
181
182
  private void updateTheta(double[] thetaDiff) {
    double[] newTheta = new double[theta.getLength()];
    for (int j = 0; j < theta.getLength(); j++) {
      newTheta[j] = theta.get(j) - thetaDiff[j] * alpha;
    }
    theta = new DenseDoubleVector(newTheta);
  }
View Full Code Here
235
236
237
238
239
240
241
242
243
  private void broadcastVector(
      BSPPeer<VectorWritable, DoubleWritable, VectorWritable, DoubleWritable, VectorWritable> peer,
      double[] vector) throws IOException {
    for (String peerName : peer.getAllPeerNames()) {
      if (!peerName.equals(peer.getPeerName())) { // avoid sending to oneself
        peer.send(peerName, new VectorWritable(new DenseDoubleVector(vector)));
      }
    }
  }
View Full Code Here
277
278
279
280
281
282
283
284
285
286
287
          BSPPeer<VectorWritable, DoubleWritable, VectorWritable, DoubleWritable, VectorWritable> peer)
      throws IOException, SyncException, InterruptedException {
    if (theta == null) {
      if (master) {
        int size = getXSize(peer);
        theta = new DenseDoubleVector(size, peer.getConfiguration().getInt(
            INITIAL_THETA_VALUES, 1));
        broadcastVector(peer, theta.toArray());
        if (log.isDebugEnabled()) {
          log.debug(new StringBuilder(peer.getPeerName()).append(
              ": sending theta").toString());
View Full Code Here
119
120
121
122
123
124
125
126
127
128
129
        double[] vals = new double[tokens.length];
        for (int i = 0; i < tokens.length; ++i) {
          vals[i] = Double.parseDouble(tokens[i]);
        }
        writer.append(new LongWritable(), new VectorWritable(
            new DenseDoubleVector(vals)));
      }
      writer.close();
      br.close();
    } catch (IOException e1) {
      e1.printStackTrace();
View Full Code Here
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
    regression.setMomemtumWeight(0.1);

    int iterations = 100;
    for (int i = 0; i < iterations; ++i) {
      for (int j = 0; j < instances.length; ++j) {
        regression.trainOnline(new DenseDoubleVector(instances[j]));
      }
    }

    double relativeError = 0;
    for (int i = 0; i < instances.length; ++i) {
      DoubleVector test = new DenseDoubleVector(instances[i]);
      double expected = test.get(test.getDimension() - 1);
      test = test.slice(test.getDimension() - 1);
      double actual = regression.getOutput(test).get(0);
      relativeError += Math.abs((expected - actual) / expected);
    }

    relativeError /= instances.length;
View Full Code Here
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
    regression.setMomemtumWeight(0.1);
    regression.setRegularizationWeight(0.05);
    int iterations = 2000;
    for (int i = 0; i < iterations; ++i) {
      for (double[] trainingInstance : trainingInstances) {
        regression.trainOnline(new DenseDoubleVector(trainingInstance));
      }
    }

    double relativeError = 0.0;
    // calculate the error on test instance
    for (double[] testInstance : testInstances) {
      DoubleVector instance = new DenseDoubleVector(testInstance);
      double expected = instance.get(instance.getDimension() - 1);
      instance = instance.slice(instance.getDimension() - 1);
      double actual = regression.getOutput(instance).get(0);
      if (expected == 0) {
        expected = 0.0000001;
      }
      relativeError += Math.abs((expected - actual) / expected);
View Full Code Here
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
    regression.setMomemtumWeight(0.1);
    regression.setRegularizationWeight(0.1);
    int iterations = 1000;
    for (int i = 0; i < iterations; ++i) {
      for (double[] trainingInstance : trainingInstances) {
        regression.trainOnline(new DenseDoubleVector(trainingInstance));
      }
    }

    double errorRate = 0;
    // calculate the error on test instance
    for (double[] testInstance : testInstances) {
      DoubleVector instance = new DenseDoubleVector(testInstance);
      double expected = instance.get(instance.getDimension() - 1);
      DoubleVector features = instance.slice(instance.getDimension() - 1);
      double actual = regression.getOutput(features).get(0);
      if (actual < 0.5 && expected >= 0.5 || actual >= 0.5 && expected < 0.5) {
        ++errorRate;
      }
View Full Code Here
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
    matrices[0] = new DenseDoubleMatrix(5, 3, 0.5);
    matrices[1] = new DenseDoubleMatrix(1, 6, 0.5);
    ann.setWeightMatrices(matrices);

    double[] arr = new double[] { 0, 1 };
    DoubleVector training = new DenseDoubleVector(arr);
    DoubleVector result = ann.getOutput(training);
    assertEquals(1, result.getDimension());
    // assertEquals(3, result.get(0), 0.000001);

    // second network
    SmallLayeredNeuralNetwork ann2 = new SmallLayeredNeuralNetwork();
    ann2.addLayer(2, false, FunctionFactory.createDoubleFunction("Sigmoid"));
    ann2.addLayer(3, false, FunctionFactory.createDoubleFunction("Sigmoid"));
    ann2.addLayer(1, true, FunctionFactory.createDoubleFunction("Sigmoid"));
    ann2.setCostFunction(FunctionFactory
        .createDoubleDoubleFunction("SquaredError"));
    ann2.setLearningRate(0.3);
    // intentionally initialize all weights to 0.5
    DoubleMatrix[] matrices2 = new DenseDoubleMatrix[2];
    matrices2[0] = new DenseDoubleMatrix(3, 3, 0.5);
    matrices2[1] = new DenseDoubleMatrix(1, 4, 0.5);
    ann2.setWeightMatrices(matrices2);

    double[] test = { 0, 0 };
    double[] result2 = { 0.807476 };

    DoubleVector vec = ann2.getOutput(new DenseDoubleVector(test));
    assertArrayEquals(result2, vec.toArray(), 0.000001);

    SmallLayeredNeuralNetwork ann3 = new SmallLayeredNeuralNetwork();
    ann3.addLayer(2, false, FunctionFactory.createDoubleFunction("Sigmoid"));
    ann3.addLayer(3, false, FunctionFactory.createDoubleFunction("Sigmoid"));
    ann3.addLayer(1, true, FunctionFactory.createDoubleFunction("Sigmoid"));
    ann3.setCostFunction(FunctionFactory
        .createDoubleDoubleFunction("SquaredError"));
    ann3.setLearningRate(0.3);
    // intentionally initialize all weights to 0.5
    DoubleMatrix[] initMatrices = new DenseDoubleMatrix[2];
    initMatrices[0] = new DenseDoubleMatrix(3, 3, 0.5);
    initMatrices[1] = new DenseDoubleMatrix(1, 4, 0.5);
    ann3.setWeightMatrices(initMatrices);

    double[] instance = { 0, 1 };
    DoubleVector output = ann3.getOutput(new DenseDoubleVector(instance));
    assertEquals(0.8315410, output.get(0), 0.000001);
  }
View Full Code Here
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
    int iterations = 50000; // iteration should be set to a very large number
    double[][] instances = { { 0, 1, 1 }, { 0, 0, 0 }, { 1, 0, 1 }, { 1, 1, 0 } };
    for (int i = 0; i < iterations; ++i) {
      DoubleMatrix[] matrices = null;
      for (int j = 0; j < instances.length; ++j) {
        matrices = ann.trainByInstance(new DenseDoubleVector(instances[j
            % instances.length]));
        ann.updateWeightMatrices(matrices);
      }
    }

    for (int i = 0; i < instances.length; ++i) {
      DoubleVector input = new DenseDoubleVector(instances[i]).slice(2);
      // the expected output is the last element in array
      double result = instances[i][2];
      double actual = ann.getOutput(input).get(0);
      if (result < 0.5 && actual >= 0.5 || result >= 0.5 && actual < 0.5) {
        Log.info("Neural network failes to lear the XOR.");
      }
    }

    // write model into file and read out
    String modelPath = "/tmp/testSmallLayeredNeuralNetworkXORLocal";
    ann.setModelPath(modelPath);
    try {
      ann.writeModelToFile();
    } catch (IOException e) {
      e.printStackTrace();
    }
    SmallLayeredNeuralNetwork annCopy = new SmallLayeredNeuralNetwork(modelPath);
    // test on instances
    for (int i = 0; i < instances.length; ++i) {
      DoubleVector input = new DenseDoubleVector(instances[i]).slice(2);
      // the expected output is the last element in array
      double result = instances[i][2];
      double actual = annCopy.getOutput(input).get(0);
      if (result < 0.5 && actual >= 0.5 || result >= 0.5 && actual < 0.5) {
        Log.info("Neural network failes to lear the XOR.");
View Full Code Here
TOP

Related Classes of org.apache.hama.ml.math.DenseDoubleVector

  • org.apache.hama.examples.NeuralNetwork
  • org.apache.hama.examples.NeuralNetworkTest
  • org.apache.hama.ml.ann.AutoEncoder
  • org.apache.hama.ml.ann.SmallLayeredNeuralNetwork
  • org.apache.hama.ml.ann.TestAutoEncoder
  • org.apache.hama.ml.ann.TestSmallLayeredNeuralNetwork
  • org.apache.hama.ml.kmeans.KMeansBSP
  • org.apache.hama.ml.perception.SmallMultiLayerPerceptron
  • org.apache.hama.ml.perception.TestSmallMultiLayerPerceptron
  • org.apache.hama.ml.regression.GradientDescentBSP
    • Copyright © 2018 www.massapicom. All rights reserved.
    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.