Java Deep Learning Cookbook

How to do it...

1. Create an array to assign weights to minor labels:

INDArray weightsArray = Nd4j.create(new double[]{0.35, 0.65});

2. Modify OutPutLayer to evenly balance the labels in the dataset:

new OutputLayer.Builder(new LossMCXENT(weightsArray)).nIn(incomingConnectionCount).nOut(labelCount).activation(Activation.SOFTMAX)).build();

3. Initialize the neural network and add the training listeners:

MultiLayerConfiguration configuration = builder.build();   MultiLayerNetwork multiLayerNetwork = new MultiLayerNetwork(configuration); multiLayerNetwork.init(); multiLayerNetwork.setListeners(new ScoreIterationListener(iterationCount));

4. Add the DL4J UI Maven dependency to analyze the training process:

<dependency> <groupId>org.deeplearning4j</groupId> ...