library(mxnet)
    train_data <- as.data.frame(Dataset[,-17])
    train_data <- as.data.frame(Dataset[,-1])
    train_data <- as.matrix(data)
    label <- as.array(round(Dataset$Mean_TRS, digits=2))
    test_data <- train_data[sample(1:nrow(train_data), 50,
                              replace=FALSE),]

    Model <- mx.mlp(data, label, hidden_node=c(128,64), out_node=2, activation="relu", out_activation="softmax",
                    num.round=100, array.batch.size=15, learning.rate=0.07, momentum=0.9, device=mx.cpu())

    Model = mx.mlp(
      data = as.matrix(train_data),
      label = label,  #as.numeric(ifelse(label == 2, 0, 1)), # Replace classes with 0 and 1
      hidden_node = 10,
      out_node = 2,
      out_activation = "softmax",
      learning.rate = 0.01,
      num.round = 50,
      array.layout = "rowmajor", # get rid of the warning
      eval.metric = mx.metric.accuracy, # set Accuracy as a metric
      momentum=0.9
    )

preds = as.data.frame(predict(Model, test_data, array.layout = "rowmajor"))