Examples of com.github.neuralnetworks.calculation.neuronfunctions.ConnectionCalculatorFullyConnected

Package com.github.neuralnetworks.calculation.neuronfunctions

Examples of com.github.neuralnetworks.calculation.neuronfunctions.ConnectionCalculatorFullyConnected

com.github.neuralnetworks.calculation.neuronfunctions.ConnectionCalculatorFullyConnected
Default implementation of Connection calculator for fully connected layers Biases are also added After all the input functions are calculated there is a list of activation functions that can be applied to the result This class differs from LayerCalculatorImpl in the fact that LayerCalculatorImpl traverses the graph of layers, where ConnectionCalculatorImpl only deals with the connections passed as parameter !!! Important !!! The results of the calculations are represented as matrices (Matrix). This is done, because it is assumed that implementations will provide a way for calculating many input results at once. Each column of the matrix represents a single input. For example if the network is trained to classify MNIST images, each column of the input matrix will represent single MNIST image.


    public static AparapiCDTrainer cdSoftReLUTrainer(RBM rbm, TrainingInputProvider trainingSet, TrainingInputProvider testingSet, OutputError error, NNRandomInitializer rand, float learningRate, float momentum, float l1weightDecay, float l2weightDecay, int gibbsSampling, boolean isPersistentCD) {
  rbm.setLayerCalculator(NNFactory.rbmSoftReluSoftRelu(rbm));


  RBMLayerCalculator lc = NNFactory.rbmSigmoidSigmoid(rbm);
  ConnectionCalculatorFullyConnected cc = (ConnectionCalculatorFullyConnected) lc.getConnectionCalculator(rbm.getInputLayer());
  cc.addPreTransferFunction(new BernoulliDistribution());


  return new AparapiCDTrainer(rbmProperties(rbm, lc, trainingSet, testingSet, error, rand, learningRate, momentum, l1weightDecay, l2weightDecay, gibbsSampling, isPersistentCD));
    }

View Full Code Here


    public static AparapiCDTrainer cdSigmoidTrainer(RBM rbm, TrainingInputProvider trainingSet, TrainingInputProvider testingSet, OutputError error, NNRandomInitializer rand, float learningRate, float momentum, float l1weightDecay, float l2weightDecay, int gibbsSampling, boolean isPersistentCD) {
  rbm.setLayerCalculator(NNFactory.rbmSigmoidSigmoid(rbm));


  RBMLayerCalculator lc = NNFactory.rbmSigmoidSigmoid(rbm);
  ConnectionCalculatorFullyConnected cc = (ConnectionCalculatorFullyConnected) lc.getConnectionCalculator(rbm.getInputLayer());
  cc.addPreTransferFunction(new BernoulliDistribution());


  return new AparapiCDTrainer(rbmProperties(rbm, lc, trainingSet, testingSet, error, rand, learningRate, momentum, l1weightDecay, l2weightDecay, gibbsSampling, isPersistentCD));
    }

View Full Code Here


  Matrix bcg = bc.getConnectionGraph();
  bcg.set(0.1f, 0, 0);
  bcg.set(0.2f, 1, 0);


  ConnectionCalculatorFullyConnected aws = new AparapiWeightedSumConnectionCalculator();


  List<Connections> connections = new ArrayList<>();
  connections.add(c1);


  ValuesProvider vp = new ValuesProvider();
  vp.addValues(c1.getInputLayer(), i1);
  vp.addValues(ol, o);


  aws.calculate(connections, vp, ol);


  // most simple case
  assertEquals(14, o.get(0, 0), 0);
  assertEquals(32, o.get(0, 1), 0);
  assertEquals(32, o.get(1, 0), 0);
  assertEquals(77, o.get(1, 1), 0);
  Util.fillArray(o.getElements(), 0);


  // with bias
  connections = new ArrayList<>();
  connections.add(c1);
  connections.add(bc);


  vp = new ValuesProvider();
  vp.addValues(c1.getInputLayer(), i1);
  vp.addValues(ol, o);


  aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  assertEquals(14.1, o.get(0, 0), 0.01);
  assertEquals(32.1, o.get(0, 1), 0.01);
  assertEquals(32.2, o.get(1, 0), 0.01);
  assertEquals(77.2, o.get(1, 1), 0.01);
  Util.fillArray(o.getElements(), 0);


  // combined layers
  connections = new ArrayList<>();
  connections.add(c1);
  connections.add(c2);
  connections.add(bc);


  vp = new ValuesProvider();
  vp.addValues(c1.getInputLayer(), i1);
  vp.addValues(c2.getInputLayer(), i2);
  vp.addValues(ol, o);


  aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  assertEquals(28.1, o.get(0, 0), 0.01);
  assertEquals(64.1, o.get(0, 1), 0.01);
  assertEquals(64.2, o.get(1, 0), 0.01);
  assertEquals(154.2, o.get(1, 1), 0.01);

View Full Code Here


  Matrix bcg = bc.getConnectionGraph();
  bcg.set(0.1f, 0, 0);
  bcg.set(0.2f, 1, 0);


  ConnectionCalculatorFullyConnected aws = new AparapiWeightedSumConnectionCalculator();


  List<Connections> connections = new ArrayList<>();
  connections.add(c1);


  ValuesProvider vp = new ValuesProvider();
  vp.addValues(c1.getOutputLayer(), i1);
  vp.addValues(ol, o);


  aws.calculate(connections, vp, ol);


  // most simple case
  assertEquals(14, o.get(0, 0), 0);
  assertEquals(32, o.get(0, 1), 0);
  assertEquals(32, o.get(1, 0), 0);
  assertEquals(77, o.get(1, 1), 0);
  Util.fillArray(o.getElements(), 0);


  // with bias
  connections = new ArrayList<>();
  connections.add(c1);
  connections.add(bc);


  vp = new ValuesProvider();
  vp.addValues(c1.getOutputLayer(), i1);
  vp.addValues(ol, o);


  aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  assertEquals(14.1, o.get(0, 0), 0.01);
  assertEquals(32.1, o.get(0, 1), 0.01);
  assertEquals(32.2, o.get(1, 0), 0.01);
  assertEquals(77.2, o.get(1, 1), 0.01);
  Util.fillArray(o.getElements(), 0);


  // combined layers
  connections = new ArrayList<>();
  connections.add(c1);
  connections.add(c2);
  connections.add(bc);


  vp = new ValuesProvider();
  vp.addValues(c1.getOutputLayer(), i1);
  vp.addValues(c2.getOutputLayer(), i2);
  vp.addValues(ol, o);


  aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  assertEquals(28.1, o.get(0, 0), 0.01);
  assertEquals(64.1, o.get(0, 1), 0.01);
  assertEquals(64.2, o.get(1, 0), 0.01);
  assertEquals(154.2, o.get(1, 1), 0.01);

View Full Code Here

  // train
  bpt.train();


  // add softmax function
  LayerCalculatorImpl lc = (LayerCalculatorImpl) mlp.getLayerCalculator();
  ConnectionCalculatorFullyConnected cc = (ConnectionCalculatorFullyConnected) lc.getConnectionCalculator(mlp.getOutputLayer());
  cc.addActivationFunction(new SoftmaxFunction());


  // test
  bpt.test();


  assertEquals(0, bpt.getOutputError().getTotalNetworkError(), 0.1);

View Full Code Here

    if (outputCC != null && nn.getOutputLayer() == l) {
        lc.addConnectionCalculator(l, outputCC);
    } else if (Util.isConvolutional(l)) {
        lc.addConnectionCalculator(l, new ConnectionCalculatorConv());
    } else {
        lc.addConnectionCalculator(l, new ConnectionCalculatorFullyConnected());
    }
      } else {
    lc.addConnectionCalculator(l, new ConstantConnectionCalculator());
      }
  }

View Full Code Here

  i1.set(3, 2, 0);
  i1.set(4, 0, 1);
  i1.set(5, 1, 1);
  i1.set(6, 2, 1);


  ConnectionCalculatorFullyConnected aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  // most simple case
  Matrix o = vp.get(nn.getOutputLayer());
  assertEquals(14, o.get(0, 0), 0);
  assertEquals(32, o.get(0, 1), 0);
  assertEquals(32, o.get(1, 0), 0);
  assertEquals(77, o.get(1, 1), 0);


  // with bias
  connections = new ArrayList<>();
  connections.add(c1);
  connections.add(bc);


  nn = new NeuralNetworkImpl();
  nn.addConnections(connections.toArray(new Connections[connections.size()]));
  vp = TensorFactory.tensorProvider(nn, 2, true);


  i1 = vp.get(nn.getInputLayer());
  i1.set(1, 0, 0);
  i1.set(2, 1, 0);
  i1.set(3, 2, 0);
  i1.set(4, 0, 1);
  i1.set(5, 1, 1);
  i1.set(6, 2, 1);


  aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  o = vp.get(nn.getOutputLayer());
  assertEquals(14.1, o.get(0, 0), 0.01);
  assertEquals(32.1, o.get(0, 1), 0.01);
  assertEquals(32.2, o.get(1, 0), 0.01);
  assertEquals(77.2, o.get(1, 1), 0.01);


  // combined layers
  connections = new ArrayList<>();
  connections.add(c1);
  connections.add(c2);
  connections.add(bc);
  nn = new NeuralNetworkImpl();
  nn.addConnections(connections.toArray(new Connections[connections.size()]));
  vp = TensorFactory.tensorProvider(nn, 2, true);


  i1 = vp.get(il1);
  i1.set(1, 0, 0);
  i1.set(2, 1, 0);
  i1.set(3, 2, 0);
  i1.set(4, 0, 1);
  i1.set(5, 1, 1);
  i1.set(6, 2, 1);


  Matrix i2 = vp.get(il2);
  i2.set(1, 0, 0);
  i2.set(2, 1, 0);
  i2.set(3, 2, 0);
  i2.set(4, 0, 1);
  i2.set(5, 1, 1);
  i2.set(6, 2, 1);


  aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  o = vp.get(nn.getOutputLayer());
  assertEquals(28.1, o.get(0, 0), 0.01);
  assertEquals(64.1, o.get(0, 1), 0.01);
  assertEquals(64.2, o.get(1, 0), 0.01);

View Full Code Here


  Matrix bcg = bc.getWeights();
  bcg.set(0.1f, 0, 0);
  bcg.set(0.2f, 1, 0);


  ConnectionCalculatorFullyConnected aws = new AparapiWeightedSumConnectionCalculator();


  List<Connections> connections = new ArrayList<>();
  connections.add(c1);
  NeuralNetworkImpl nn = new NeuralNetworkImpl();
  nn.addConnections(connections.toArray(new Connections[connections.size()]));
  ValuesProvider vp = TensorFactory.tensorProvider(nn, 2, true);


  Matrix i1 = vp.get(il1);
  i1.set(1, 0, 0);
  i1.set(2, 1, 0);
  i1.set(3, 2, 0);
  i1.set(4, 0, 1);
  i1.set(5, 1, 1);
  i1.set(6, 2, 1);


  aws.calculate(connections, vp, ol);


  // most simple case
  Matrix o = vp.get(ol);
  assertEquals(14, o.get(0, 0), 0);
  assertEquals(32, o.get(0, 1), 0);
  assertEquals(32, o.get(1, 0), 0);
  assertEquals(77, o.get(1, 1), 0);


  // with bias
  connections = new ArrayList<>();
  connections.add(c1);
  connections.add(bc);
  nn = new NeuralNetworkImpl();
  nn.addConnections(connections.toArray(new Connections[connections.size()]));
  vp = TensorFactory.tensorProvider(nn, 2, true);
  i1 = vp.get(il1);
  i1.set(1, 0, 0);
  i1.set(2, 1, 0);
  i1.set(3, 2, 0);
  i1.set(4, 0, 1);
  i1.set(5, 1, 1);
  i1.set(6, 2, 1);


  aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  o = vp.get(ol);
  assertEquals(14.1, o.get(0, 0), 0.01);
  assertEquals(32.1, o.get(0, 1), 0.01);
  assertEquals(32.2, o.get(1, 0), 0.01);
  assertEquals(77.2, o.get(1, 1), 0.01);


  // combined layers
  connections = new ArrayList<>();
  connections.add(c1);
  connections.add(c2);
  connections.add(bc);
  nn = new NeuralNetworkImpl();
  nn.addConnections(connections.toArray(new Connections[connections.size()]));
  vp = TensorFactory.tensorProvider(nn, 2, true);


  i1 = vp.get(il1);
  i1.set(1, 0, 0);
  i1.set(2, 1, 0);
  i1.set(3, 2, 0);
  i1.set(4, 0, 1);
  i1.set(5, 1, 1);
  i1.set(6, 2, 1);


  Matrix i2 = vp.get(il2);
  i2.set(1, 0, 0);
  i2.set(2, 1, 0);
  i2.set(3, 2, 0);
  i2.set(4, 0, 1);
  i2.set(5, 1, 1);
  i2.set(6, 2, 1);


  aws = new AparapiWeightedSumConnectionCalculator();
  aws.calculate(connections, vp, ol);


  o = vp.get(ol);
  assertEquals(28.1, o.get(0, 0), 0.01);
  assertEquals(64.1, o.get(0, 1), 0.01);
  assertEquals(64.2, o.get(1, 0), 0.01);

View Full Code Here

  float dropoutRate = properties.getParameter(Constants.DROPOUT_RATE);


  if (dropoutRate > 0) {
      LayerCalculatorImpl lc = (LayerCalculatorImpl) nn.getLayerCalculator();
      nn.getConnections().stream().filter(c -> c instanceof FullyConnected && c.getInputLayer() != nn.getInputLayer() && !Util.isBias(c.getInputLayer())).forEach(c -> {
    ConnectionCalculatorFullyConnected cc = (ConnectionCalculatorFullyConnected) lc.getConnectionCalculator(c.getOutputLayer());
    cc.setDropoutRate(dropoutRate);
      });


      addEventListener(this);
  }
    }

View Full Code Here

      if (dropoutRate > 0) {
    NeuralNetwork nn = getNeuralNetwork();


    LayerCalculatorImpl lc = (LayerCalculatorImpl) nn.getLayerCalculator();
    nn.getConnections().stream().filter(c -> c instanceof FullyConnected && c.getInputLayer() != nn.getInputLayer() && !Util.isBias(c.getInputLayer())).forEach(c -> {
        ConnectionCalculatorFullyConnected cc = (ConnectionCalculatorFullyConnected) lc.getConnectionCalculator(c.getOutputLayer());
        cc.setDropoutRate(0);
        FullyConnected fc = (FullyConnected) c;
        fc.getWeights().forEach(i -> fc.getWeights().getElements()[i] = fc.getWeights().getElements()[i] * (1 - dropoutRate));
    });
      }
  }

View Full Code Here

0 1

TOP

Related Classes of com.github.neuralnetworks.calculation.neuronfunctions.ConnectionCalculatorFullyConnected

com.github.neuralnetworks.architecture.Connections

com.github.neuralnetworks.architecture.GraphConnections

com.github.neuralnetworks.architecture.types.NNFactory

com.github.neuralnetworks.calculation.ConnectionCalculator

com.github.neuralnetworks.samples.test.IrisTest

com.github.neuralnetworks.tensor.Matrix

com.github.neuralnetworks.tensor.Tensor

com.github.neuralnetworks.test.FFNNTest

com.github.neuralnetworks.training.backpropagation.BackPropagationTrainer

com.github.neuralnetworks.training.TrainerFactory

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.