Scala Machine Learning Projects

Step 8 - Training the LSTM network

Now we will start training the LSTM network. However, before getting started, let's try to define some variables to keep track of the training's performance:

val testLosses = ArrayBuffer[Float]() 
val testAccuracies = ArrayBuffer[Float]() 
val trainLosses = ArrayBuffer[Float]() 
val trainAccuracies = ArrayBuffer[Float]()     

Then, we start performing the training steps with batch_size iterations at each loop:

var step = 1 
while (step * batchSize <= trainingIters) { 
    val (batchTrainData, batchTrainLabel) = { 
        val idx = ((step - 1) * batchSize) % trainingDataCount 
        if (idx + batchSize <= trainingDataCount) { 
          val datas = trainData.drop(idx).take(batchSize) 
          val labels = trainLabels.drop(idx).take(batchSize) (datas, ...