NNET

val conf = new NeuralNetConfiguration.Builder().regularization(true).l2(0.001).weightInit(WeightInit.RELU)
  .updater(new Adam(0.01, 0.9, .98, 1e-8)).
  optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT).
  convolutionMode(ConvolutionMode.Strict).list().
  layer(0, new ConvolutionLayer.Builder(7, 7).nIn(1).stride(1, 1)
    .nOut(25).activation(Activation.IDENTITY).build())
  .layer(1, new SubsamplingLayer.Builder(PoolingType.MAX).
    kernelSize(2, 2).stride(2, 2).build()).
  layer(2, new ConvolutionLayer.Builder(6, 6).stride(1, 1).nOut(64).
    activation(Activation.IDENTITY).build()).
  layer(3, new SubsamplingLayer.Builder(PoolingType.MAX).
    kernelSize(2, 2).stride(2, 2).build()).
  layer(4, new ConvolutionLayer.Builder(5, 5).stride(1, 1).nOut(120).
    activation(Activation.IDENTITY).build()).
  layer(5, new SubsamplingLayer.Builder(PoolingType.MAX).
    kernelSize(2, 2).stride(2, 2).build()).
  layer(6, new DenseLayer.Builder().activation(Activation.RELU).nOut(1000).build())
  .layer(7, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD).
    nOut(2).activation(Activation.SOFTMAX).build())
  .setInputType(InputType.convolutional(128, 128, 1)).
  backprop(true).pretrain(false).build()
val labelGenerator = new ParentPathLabelGenerator()
val imageResizer = new MultiImageTransform(new ScaleImageTransform(128, 128))
val imageRecordReaderTrain = new ImageRecordReader(128, 128, 1, labelGenerator)
imageRecordReaderTrain.initialize(data(0),imageResizer)
val scaler = new ImagePreProcessingScaler(0, 1)
val imageRecordReaderTest = new ImageRecordReader(128, 128, 1, labelGenerator)
imageRecordReaderTest.initialize(data(1),imageResizer)
val trainIter = new RecordReaderDataSetIterator(imageRecordReaderTrain, 10, 1, 2)
scaler.fit(trainIter)
trainIter.setPreProcessor(scaler)
val testIter = new RecordReaderDataSetIterator(imageRecordReaderTest, 10, 1, 2)
testIter.setPreProcessor(scaler)
val esConf = new EarlyStoppingConfiguration.Builder().
  epochTerminationConditions(new ScoreImprovementEpochTerminationCondition(10))
  .scoreCalculator(new DataSetLossCalculator(testIter, true))
  .evaluateEveryNEpochs(1)
  .modelSaver(new LocalFileModelSaver(rb.getString("model_loc"))).build()
val model = new MultiLayerNetwork(conf)
model.init()
val uiServer = UIServer.getInstance()
val statsStorage = new InMemoryStatsStorage
uiServer.attach(statsStorage)
model.setListeners(new ScoreIterationListener(10))
model.setListeners(new StatsListener(statsStorage))
val esTrainer = new EarlyStoppingTrainer(esConf, model, trainIter)
val esResult = esTrainer.fit()
logger.info("Termination Reason {}", esResult.getTerminationReason)
logger.info("Termination Details {}", esResult.getTerminationDetails)
logger.info("Epochs{}", esResult.getTotalEpochs)
logger.info("Best Epoch {}", esResult.getBestModelEpoch)
logger.info("Score at best epoch{}", esResult.getBestModelScore)
testIter.reset()
logger.info("Evaluating test score {}",esResult.getBestModel.evaluate(testIter))