.l2(1e-5) .weightInit(WeightInit.XAVIER) .gradientNormalization(GradientNormalization.ClipElementWiseAbsoluteValue).gradientNormalizationThreshold(1.0) .list() .layer(0, new LSTM.Builder().nIn(vectorSize).nOut(256)
.updater(Updater.RMSPROP).regularization(true).l2(1e-5) .weightInit(WeightInit.XAVIER) .gradientNormalization(GradientNormalization.ClipElementWiseAbsoluteValue) .gradientNormalizationThreshold(1.0).learningRate(0.0018).list() .layer(0,
.iterations(iterations).seed(12345l).updater(Updater.SGD).regularization(true) .l2(1e-5).weightInit(WeightInit.RELU) .gradientNormalization(GradientNormalization.ClipElementWiseAbsoluteValue) .gradientNormalizationThreshold(1.0).learningRate(learningRate).list() .layer(0,
.updater(new Nesterovs(new StepSchedule(ScheduleType.ITERATION, 1e-2, 0.1, 100000), 0.9)) .biasUpdater(new Nesterovs(new StepSchedule(ScheduleType.ITERATION, 2e-2, 0.1, 100000), 0.9)) .gradientNormalization(GradientNormalization.RenormalizeL2PerLayer) // normalize to prevent vanishing or exploding gradients .l2(5 * 1e-4)
builder.rmsDecay(rmsDecay.getValue(values)); if (gradientNormalization != null) builder.gradientNormalization(gradientNormalization.getValue(values)); if (gradientNormalizationThreshold != null) builder.gradientNormalizationThreshold(gradientNormalizationThreshold.getValue(values));
.updater(Updater.NESTEROVS).learningRate(1e-2).biasLearningRate(1e-2 * 2).regularization(true) .convolutionMode(ConvolutionMode.Same) .gradientNormalization(GradientNormalization.RenormalizeL2PerLayer) // normalize to prevent vanishing or exploding gradients .trainingWorkspaceMode(WorkspaceMode.SINGLE).inferenceWorkspaceMode(WorkspaceMode.SINGLE) .dropOut(0.5).l2(5 * 1e-4).miniBatch(false)
.updater(new Nesterovs(new StepSchedule(ScheduleType.ITERATION, 0.1, 0.1, 100000), 0.9)) .biasUpdater(new Nesterovs(new StepSchedule(ScheduleType.ITERATION, 0.2, 0.1, 100000), 0.9)) .gradientNormalization(GradientNormalization.RenormalizeL2PerLayer) // normalize to prevent vanishing or exploding gradients
.l2(aTraits.getL2()) .weightInit(aTraits.getWeightInit()) .gradientNormalization(aTraits.getGradientNormalization()) .gradientNormalizationThreshold(aTraits.getGradientNormalizationThreshold()) .list()
.l2(aTraits.getL2()) .weightInit(aTraits.getWeightInit()) .gradientNormalization(aTraits.getGradientNormalization()) .gradientNormalizationThreshold(aTraits.getGradientNormalizationThreshold()) .list()
.l2(1e-5) .weightInit(WeightInit.XAVIER) .gradientNormalization(GradientNormalization.ClipElementWiseAbsoluteValue) .gradientNormalizationThreshold(1.0) .list()
/** * Deliver access to the internal builder * * @return NeuralNetworkConfiguration */ public org.deeplearning4j.nn.conf.NeuralNetConfiguration.Builder builder() { Builder builder = new Builder(); // Set dist to null if Disabled was chosen as dl4j backend defaults to null builder .l1(l1) .l2(l2) .optimizationAlgo(optimizationAlgo) .seed(seed) .weightInit(weightInit) .dist(dist.getBackend()) .biasInit(biasInit) .updater(updater.getBackend()) .biasUpdater(biasUpdater.getBackend()) .dropOut(dropout.getBackend()) .miniBatch(miniBatch) .minimize(minimize) .weightNoise(weightNoise.getBackend()) .gradientNormalization(gradientNormalization.getBackend()) .gradientNormalizationThreshold(gradientNormalizationThreshold) .inferenceWorkspaceMode(inferenceWorkspaceMode) .trainingWorkspaceMode(trainingWorkspaceMode); builder.setPretrain(pretrain); return builder; }