.build()) .layer(4, new DenseLayer.Builder().activation(Activation.RELU) .nOut(500).build()) .layer(5, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD) .nOut(outputNum)
public MultiLayerConfiguration conf() { MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder().trainingWorkspaceMode(workspaceMode) .inferenceWorkspaceMode(workspaceMode).seed(seed).iterations(iterations) .activation(Activation.IDENTITY).weightInit(WeightInit.XAVIER) .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT).updater(new AdaDelta()) .regularization(false).convolutionMode(ConvolutionMode.Same).list() // block 1 .layer(0, new ConvolutionLayer.Builder(new int[] {5, 5}, new int[] {1, 1}).name("cnn1") .nIn(inputShape[0]).nOut(20).activation(Activation.RELU).build()) .layer(1, new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX, new int[] {2, 2}, new int[] {2, 2}).name("maxpool1").build()) // block 2 .layer(2, new ConvolutionLayer.Builder(new int[] {5, 5}, new int[] {1, 1}).name("cnn2").nOut(50) .activation(Activation.RELU).build()) .layer(3, new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX, new int[] {2, 2}, new int[] {2, 2}).name("maxpool2").build()) // fully connected .layer(4, new DenseLayer.Builder().name("ffn1").activation(Activation.RELU).nOut(500).build()) // output .layer(5, new OutputLayer.Builder(LossFunctions.LossFunction.MCXENT).name("output") .nOut(numLabels).activation(Activation.SOFTMAX) // radial basis function required .build()) .setInputType(InputType.convolutionalFlat(inputShape[2], inputShape[1], inputShape[0])) .backprop(true).pretrain(false).build(); return conf; }
ComputationGraphConfiguration.GraphBuilder assemble(int numInputs, int numHiddenNodes, int numLayers, String baseLayer, int startingIndex) { assert numHiddenNodes > 0 : "model capacity is too small. At least some hidden nodes must be created."; WeightInit WEIGHT_INIT = WeightInit.XAVIER; float reduction = 1f; int minimum = (int) (numHiddenNodes * Math.pow(reduction, numLayers)); assert minimum > 2 : "Too much reduction, not enough outputs: "; int numIn = numInputs; int numOut = (int) (numHiddenNodes * modelCapacity); String previousLayerName; String lastDenseLayerName = "no layers"; for (int i = startingIndex; i < startingIndex + numLayers; i++) { // System.out.printf("layer %d numIn=%d numOut=%d%n", i, numIn, numOut); lastDenseLayerName = "dense" + i; previousLayerName = i == startingIndex ? baseLayer : "dense" + (i - 1); numOut = (int) (numHiddenNodes * Math.pow(reductionRate, i) * modelCapacity); build.addLayer(lastDenseLayerName, new DenseLayer.Builder().nIn(numIn).nOut(numOut) .weightInit(WEIGHT_INIT) .activation("relu").learningRateDecayPolicy(learningRatePolicy).epsilon(LAYER_EPSILON) .build(), previousLayerName); numIn = numOut; } this.numOutputs = numOut; this.lastLayerName = lastDenseLayerName; return build; }
@Override public ComputationGraph init() { int embeddingSize = 128; ComputationGraphConfiguration.GraphBuilder graph = graphBuilder("input1"); graph.addInputs("input1").setInputTypes(InputType.convolutional(inputShape[2], inputShape[1], inputShape[0])) // Logits .addLayer("bottleneck", new DenseLayer.Builder().nIn(5376).nOut(embeddingSize).build(), "avgpool") // Embeddings .addVertex("embeddings", new L2NormalizeVertex(new int[] {1}, 1e-10), "bottleneck") // Output .addLayer("outputLayer", new CenterLossOutputLayer.Builder() .lossFunction(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD) .activation(Activation.SOFTMAX).alpha(0.9).lambda(1e-4) .nIn(embeddingSize).nOut(numClasses).build(), "embeddings") .setOutputs("outputLayer").backprop(true).pretrain(false); ComputationGraphConfiguration conf = graph.build(); ComputationGraph model = new ComputationGraph(conf); model.init(); return model; }
public static void main(String[] args) throws Exception { //Define a simple ComputationGraph: ComputationGraphConfiguration conf = new NeuralNetConfiguration.Builder() .weightInit(WeightInit.XAVIER) .updater(new Nesterovs(0.01, 0.9)) .graphBuilder() .addInputs("in") .addLayer("layer0", new DenseLayer.Builder().nIn(4).nOut(3).activation(Activation.TANH).build(), "in") .addLayer("layer1", new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD).activation(Activation.SOFTMAX).nIn(3).nOut(3).build(), "layer0") .setOutputs("layer1") .backprop(true).pretrain(false).build(); ComputationGraph net = new ComputationGraph(conf); net.init(); //Save the model File locationToSave = new File("model/MyComputationGraph.zip"); //Where to save the network. Note: the file is in .zip format - can be opened externally boolean saveUpdater = true; //Updater: i.e., the state for Momentum, RMSProp, Adagrad etc. Save this if you want to train your network more in the future ModelSerializer.writeModel(net, locationToSave, saveUpdater); //Load the model ComputationGraph restored = ModelSerializer.restoreComputationGraph(locationToSave); System.out.println("Saved and loaded parameters are equal: " + net.params().equals(restored.params())); System.out.println("Saved and loaded configurations are equal: " + net.getConfiguration().equals(restored.getConfiguration())); }
/** Returns the network configuration, 2 hidden DenseLayers of size 50. */ private static MultiLayerConfiguration getDeepDenseLayerNetworkConfiguration() { final int numHiddenNodes = 100; return new NeuralNetConfiguration.Builder() .seed(seed) .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT) .weightInit(WeightInit.XAVIER) .updater(new Nesterovs(learningRate, 0.9)) .list() .layer(0, new DenseLayer.Builder().nIn(numInputs).nOut(numHiddenNodes) .activation(Activation.RELU).build()) .layer(1, new DenseLayer.Builder().nIn(numHiddenNodes).nOut(numHiddenNodes) .activation(Activation.RELU).build()) .layer(2, new OutputLayer.Builder(LossFunctions.LossFunction.MSE) .activation(Activation.IDENTITY) .nIn(numHiddenNodes).nOut(numOutputs).build()) .pretrain(false).backprop(true).build(); }
/** * Constructor from parsed Keras layer configuration dictionary. * * @param layerConfig dictionary containing Keras layer configuration * @param enforceTrainingConfig whether to enforce training-related configuration options * @throws InvalidKerasConfigurationException * @throws UnsupportedKerasConfigurationException */ public KerasDense(Map<String, Object> layerConfig, boolean enforceTrainingConfig) throws InvalidKerasConfigurationException, UnsupportedKerasConfigurationException { super(layerConfig, enforceTrainingConfig); this.layer = new DenseLayer.Builder().name(this.layerName).nOut(getNOutFromConfig(layerConfig)) .dropOut(this.dropout).activation(getActivationFromConfig(layerConfig)) .weightInit(getWeightInitFromConfig(layerConfig, enforceTrainingConfig)).biasInit(0.0) .l1(this.weightL1Regularization).l2(this.weightL2Regularization).build(); }
private static DenseLayer fullyConnected(String name, int out, double bias, double dropOut, Distribution dist) { return new DenseLayer.Builder().name(name).nOut(out).biasInit(bias).dropOut(dropOut).dist(dist).build(); }
public static DenseLayer fullyConnected(int in, int out, double dropOut) { return new DenseLayer.Builder().nIn(in).nOut(out).dropOut(dropOut).build(); }
private DenseLayer fullyConnected(int in, int out, double dropOut) { return new DenseLayer.Builder().nIn(in).nOut(out).dropOut(dropOut).build(); }
@Override public DenseLayer getValue(double[] values) { //Using the builder here, to get default options DenseLayer.Builder b = new DenseLayer.Builder(); setLayerOptionsBuilder(b, values); return b.build(); }
.build()) .layer(4, new DenseLayer.Builder().activation(Activation.RELU) .nOut(500).build()) .layer(5, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD) .nOut(outputNum)
.build()) .layer(4, new DenseLayer.Builder().activation(Activation.RELU) .nOut(500).build()) .layer(5, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD) .nOut(outputNum)
.gradientNormalization(GradientNormalization.ClipElementWiseAbsoluteValue) .gradientNormalizationThreshold(10) .build()) .layer(4, new LSTM.Builder() .activation(Activation.SOFTSIGN)
public static MultiLayerNetwork lenetModel() { /** * Revisde Lenet Model approach developed by ramgo2 achieves slightly above random * Reference: https://gist.github.com/ramgo2/833f12e92359a2da9e5c2fb6333351c5 **/ MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder() .seed(seed) .l2(0.005) // tried 0.0001, 0.0005 .activation(Activation.RELU) .weightInit(WeightInit.XAVIER) .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT) .updater(new Nesterovs(0.0001,0.9)) .list() .layer(0, new ConvolutionLayer.Builder(new int[]{5, 5}, new int[]{1, 1}, new int[]{0, 0}).name("cnn1") .nIn(channels).nOut(50).biasInit(0).build()) .layer(1, new SubsamplingLayer.Builder(new int[]{2,2}, new int[]{2,2}).name("maxpool1").build()) .layer(2, new ConvolutionLayer.Builder(new int[]{5,5}, new int[]{5, 5}, new int[]{1, 1}).name("cnn2") .nOut(100).biasInit(0).build()) .layer(3, new SubsamplingLayer.Builder(new int[]{2,2}, new int[]{2,2}).name("maxpool2").build()) .layer(4, new DenseLayer.Builder().nOut(500).build()) .layer(5, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD) .nOut(4) .activation(Activation.SOFTMAX) .build()) .backprop(true).pretrain(false) .setInputType(InputType.convolutional(height, width, channels)) .build(); return new MultiLayerNetwork(conf); }
.build()) .layer(1, new OutputLayer.Builder(new AutoRecLossFunction()).nIn(hiddenDim).nOut(inputDim)
public static ComputationGraphConfiguration getConf() { ComputationGraphConfiguration.GraphBuilder builder = new NeuralNetConfiguration.Builder() .seed(12345) .updater(new Adam(0.01)) .weightInit(WeightInit.RELU) .graphBuilder() .addInputs("in"); String[] poolNames = new String[ngramFilters.length]; int i = 0; for (int ngram : ngramFilters) { String filterName = String.format("ngram%d", ngram); poolNames[i] = String.format("pool%d", ngram); builder = builder.addLayer(filterName, new Convolution1DLayer.Builder() .nOut(numFilters) .kernelSize(ngram) .activation(Activation.RELU) .build(), "in") .addLayer(poolNames[i], new GlobalPoolingLayer.Builder(PoolingType.MAX).build(), filterName); i++; } return builder.addVertex("concat", new MergeVertex(), poolNames) .addLayer("predict", new DenseLayer.Builder().nOut(numClasses).dropOut(dropoutRetain) .activation(Activation.SOFTMAX).build(), "concat") .addLayer("loss", new LossLayer.Builder(LossFunctions.LossFunction.MCXENT).build(), "predict") .setOutputs("loss") .setInputTypes(InputType.recurrent(W2V_VECTOR_SIZE, 1000)) .build(); } }
public static MultiLayerConfiguration lenetModelConf() { MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder() .seed(seed) .l2(0.005) .activation(Activation.RELU) .weightInit(WeightInit.XAVIER) .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT) .updater(new Nesterovs(0.0001, 0.9)) .list() .layer(0, new ConvolutionLayer.Builder(new int[]{5, 5}, new int[]{1, 1}, new int[]{0, 0}).name("cnn1") .nIn(channels).nOut(50).biasInit(0).build()) .layer(1, new SubsamplingLayer.Builder(new int[]{2,2}, new int[]{2,2}).name("maxpool1").build()) .layer(2, new ConvolutionLayer.Builder(new int[]{5,5}, new int[]{5, 5}, new int[]{1, 1}).name("cnn2") .nOut(100).biasInit(0).build()) .layer(3, new SubsamplingLayer.Builder(new int[]{2,2}, new int[]{2,2}).name("maxpool2").build()) .layer(4, new DenseLayer.Builder().nOut(500).build()) .layer(5, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD) .nOut(4) .activation(Activation.SOFTMAX) .build()) .backprop(true).pretrain(false) .setInputType(InputType.convolutional(height, width, channels)) .build(); return conf; } public static void saveModel(FileSystem fs, Model model ) throws Exception{