RnnToFeedForward

gistfile1.txt

		// MultiLayerConfiguration
    int OutDemension = 2;
    ListBuilder listBuilder = builder.list();
		int layerIndex = 0;
		listBuilder.layer(layerIndex++, new GravesLSTM.Builder().activation(Activation.SOFTSIGN) // SOFTSIGN, TANH, RELU, SIGMOID
				.nIn(inNum).nOut(hiddenCount).build());
		listBuilder.layer(layerIndex++, new GravesLSTM.Builder().activation(Activation.SOFTSIGN)
				.nIn(hiddenCount).nOut(hiddenCount).build());
		listBuilder.layer(layerIndex++, new DenseLayer.Builder().activation(Activation.TANH)
				.nIn(hiddenCount).nOut(denseLayerHiddenCount).build());
		listBuilder.layer(layerIndex++, new DenseLayer.Builder().activation(Activation.TANH)
				.nIn(denseLayerHiddenCount).nOut(denseLayerHiddenCount / 4).build());
		// 1, RnnOutputLayer
		// listBuilder.layer(layerIndex++,
		// new RnnOutputLayer.Builder(LossFunctions.LossFunction.MSE).activation(Activation.IDENTITY).nIn(hiddenCount).nOut(OutDemension).build());
		// 2, OutputLayer
		listBuilder.layer(layerIndex++, new OutputLayer.Builder(LossFunctions.LossFunction.MSE).activation(Activation.IDENTITY)
				.nIn(denseLayerHiddenCount /4)
				.nOut(OutDemension).build());
		listBuilder.inputPreProcessor(2, new RnnToFeedForwardPreProcessor());
    
    
			// predicting
			// 1, using RnnOutputLayer
			// INDArray predicted = net.rnnTimeStep(testData.getFeatureMatrix());
			// 2, using OutputLayer
			INDArray predicted = net.output(testData.getFeatureMatrix());