Examples of tv.floe.metronome.deeplearning.neuralnetwork.layer.HiddenLayer

Package tv.floe.metronome.deeplearning.neuralnetwork.layer

Examples of tv.floe.metronome.deeplearning.neuralnetwork.layer.HiddenLayer

tv.floe.metronome.deeplearning.neuralnetwork.layer.HiddenLayer
Hidden Layer TODO: explain the role of a hidden layer in a Deep Belief Network - compared to a RBM, or a traditional hidden layer @author josh

        
        this.hiddenLayers = new HiddenLayer[ this.numberLayers ];
        // write in hidden layers
        for ( int x = 0; x < this.numberLayers; x++ ) {


          this.hiddenLayers[ x ] = new HiddenLayer( 1, 1, null); 
          this.hiddenLayers[ x ].load( is );
          
          
        }

View Full Code Here

    /* sanity check for hidden layer and inter layer dimensions */
    private void dimensionCheck() {


        for (int i = 0; i < this.numberLayers; i++) {


            HiddenLayer h = this.hiddenLayers[i];
            NeuralNetworkVectorized network = this.preTrainingLayers[i];


            //h.getW().assertSameSize(network.getW());
            MatrixUtils.assertSameLength( h.connectionWeights, network.getConnectionWeights() );
//            h.getB().assertSameSize(network.gethBias());
            MatrixUtils.assertSameLength( h.biasTerms, network.getHiddenBias() );




            if (i < this.numberLayers - 1) {




                HiddenLayer h1 = this.hiddenLayers[ i + 1 ];
                NeuralNetworkVectorized network1 = this.preTrainingLayers[ i + 1 ];


                if ( h1.getNeuronInputCount() != h.getNeuronOutputCount() ) {
                    throw new IllegalStateException("Invalid structure: hidden layer in for " + (i + 1) + " not equal to number of ins " + i);
                }


                if (network.getnHidden() != network1.getnVisible()) {
                    throw new IllegalStateException("Invalid structure: network hidden for " + (i + 1) + " not equal to number of visible " + i);

View Full Code Here

                //input_size = this.nIns;
                input_size = this.inputNeuronCount;


                // construct sigmoid_layer
                //this.sigmoidLayers[i] = new HiddenLayer(input_size, this.hiddenLayerSizes[i], null, null, rng,layer_input);
                this.hiddenLayers[ i ] = new HiddenLayer(input_size, this.hiddenLayerSizes[i], this.randomGenerator );
                this.hiddenLayers[ i ].setInput( layer_input );






            } else {


                input_size = this.hiddenLayerSizes[ i - 1 ];
                layer_input = this.hiddenLayers[i - 1].sampleHiddenGivenLastVisible();
                // construct sigmoid_layer
                this.hiddenLayers[ i ] = new HiddenLayer(input_size, this.hiddenLayerSizes[i], this.randomGenerator);
                this.hiddenLayers[ i ].setInput( layer_input );


            }


            // construct DL appropriate class for pre training layer

View Full Code Here

*/


            //getLayers()[i].setInput(currInput);
            this.preTrainingLayers[i].setInput(currInput);
            //getSigmoidLayers()[i].setInput(input);
            HiddenLayer layer = this.hiddenLayers[ i ];
            layer.setInput( this.inputTrainingData );


            if (useHiddenActivationsForwardProp) {
                //currInput = getSigmoidLayers()[i].activate(currInput);
                currInput = layer.computeOutputActivation(currInput);
            } else {
                //currInput = getLayers()[i].sampleHiddenGivenVisible(currInput).getSecond();
                currInput = this.preTrainingLayers[ i ].sampleHiddenGivenVisible(currInput).getSecond();
            }

View Full Code Here

    public Matrix predict(Matrix x) {


        Matrix input = x;


        for(int i = 0; i < this.numberLayers; i++) {
            HiddenLayer layer = this.hiddenLayers[i];
            input = layer.computeOutputActivation(input);
        }


        return this.logisticRegressionLayer.predict(input);
    }

View Full Code Here

TOP

Related Classes of tv.floe.metronome.deeplearning.neuralnetwork.layer.HiddenLayer

org.apache.commons.math3.distribution.NormalDistribution

org.apache.commons.math3.random.MersenneTwister

org.apache.mahout.math.DenseMatrix

org.apache.mahout.math.Matrix

tv.floe.metronome.deeplearning.dbn.DeepBeliefNetwork

tv.floe.metronome.deeplearning.neuralnetwork.core.BaseMultiLayerNeuralNetworkVectorized

java.io.ObjectOutputStream

java.io.ObjectInputStream

java.io.DataInput

java.io.DataOutputStream

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.