1. Load the word vector model using WordVectorSerializer:
   

WordVectors wordVectors = WordVectorSerializer.loadStaticModel(new File(WORD_VECTORS_PATH));

2. Create a sentence provider using FileLabeledSentenceProvider:

 Map<String,List<File>> reviewFilesMap = new HashMap<>();
 reviewFilesMap.put("Positive", Arrays.asList(filePositive.listFiles()));
 reviewFilesMap.put("Negative", Arrays.asList(fileNegative.listFiles()));
 LabeledSentenceProvider sentenceProvider = new FileLabeledSentenceProvider(reviewFilesMap, rndSeed); 

3. Create train iterators or test iterators using CnnSentenceDataSetIterator to load the IMDB review data:

CnnSentenceDataSetIterator iterator = new CnnSentenceDataSetIterator.Builder(CnnSentenceDataSetIterator.Format.CNN2D)
 .sentenceProvider(sentenceProvider)
 .wordVectors(wordVectors) //we mention word vectors here
 .minibatchSize(minibatchSize)
 .maxSentenceLength(maxSentenceLength) //words with length greater than this will be ignored.
 .useNormalizedWordVectors(false)
 .build();
 

4. Create a ComputationGraph configuration by adding default hyperparameters:

ComputationGraphConfiguration.GraphBuilder builder = new NeuralNetConfiguration.Builder()
 .weightInit(WeightInit.RELU)
 .activation(Activation.LEAKYRELU)
 .updater(new Adam(0.01))
 .convolutionMode(ConvolutionMode.Same) //This is important so we can 'stack' the results later
 .l2(0.0001).graphBuilder();

5. Configure layers for ComputationGraph using the addLayer() method:
   

builder.addLayer("cnn3", new ConvolutionLayer.Builder()
 .kernelSize(3,vectorSize) //vectorSize=300 for google vectors
 .stride(1,vectorSize)
 .nOut(100)
 .build(), "input");
 builder.addLayer("cnn4", new ConvolutionLayer.Builder()
 .kernelSize(4,vectorSize)
 .stride(1,vectorSize)
 .nOut(100)
 .build(), "input");
 builder.addLayer("cnn5", new ConvolutionLayer.Builder()
 .kernelSize(5,vectorSize)
 .stride(1,vectorSize)
 .nOut(100)
 .build(), "input");

6. Set the convolution mode to stack the results later:

builder.addVertex("merge", new MergeVertex(), "cnn3", "cnn4", "cnn5")

7. Create a ComputationGraph model and initialize it:
   

ComputationGraphConfiguration config = builder.build();
 ComputationGraph net = new ComputationGraph(config);
  net.init();

8. Perform the training using the fit() method:

for (int i = 0; i < numEpochs; i++) {
 net.fit(trainIterator);
 }

9. Evaluate the results:

Evaluation evaluation = net.evaluate(testIter);
System.out.println(evaluation.stats());

10. Retrieve predictions for the IMDB reviews data:

INDArray features = ((CnnSentenceDataSetIterator)testIterator).loadSingleSentence(contents);
 INDArray predictions = net.outputSingle(features);
 List<String> labels = testIterator.getLabels();
 System.out.println("

Predictions for first negative review:");
 for( int i=0; i<labels.size(); i++ ){
 System.out.println("P(" + labels.get(i) + ") = " + predictions.getDouble(i));
 }