examples/example.mnist.R

In maddin79/darch: Package for Deep Architectures and Restricted Boltzmann Machines

# MNIST example with pre-training
example.mnist <- function(dataFolder = "data/", downloadMNIST = F, ...)
{
  # Make sure to prove the correct folder if you have already downloaded the
  # MNIST data somewhere, or otherwise set downloadMNIST to TRUE
  provideMNIST(dataFolder, downloadMNIST)
  
  # Load MNIST data
  load(paste0(dataFolder, "train.RData")) # trainData, trainLabels
  load(paste0(dataFolder, "test.RData")) # testData, testLabels
  
  # only take 1000 samples, otherwise training takes increasingly long
  chosenRowsTrain <- sample(1:nrow(trainData), size=1000)
  trainDataSmall <- trainData[chosenRowsTrain,]
  trainLabelsSmall <- trainLabels[chosenRowsTrain,]
  
  darch  <- darch(trainDataSmall, trainLabelsSmall,
    rbm.numEpochs = 5,
    rbm.consecutive = F, # each RBM is trained one epoch at a time
    rbm.batchSize = 100,
    rbm.lastLayer = -1, # don't train output layer
    rbm.allData = T, # use bootstrap validation data as well for training
    rbm.errorFunction = rmseError,
    rbm.initialMomentum = .5,
    rbm.finalMomentum = .7,
    rbm.learnRate = .1,
    rbm.learnRateScale = .98,
    rbm.momentumRampLength = .8,
    rbm.numCD = 2,
    rbm.unitFunction = sigmoidUnitRbm,
    rbm.weightDecay = .001,
    layers = c(784,100,10),
    darch.batchSize = 100,
    darch.dither = T,
    darch.initialMomentum = .4,
    darch.finalMomentum = .9,
    darch.momentumRampLength = .75,
    bp.learnRate = 1,
    bp.learnRateScale = .99,
    darch.unitFunction = c(tanhUnit, softmaxUnit),
    bootstrap = T,
    darch.numEpochs = 20,
    gputools = T, # try to use gputools
    gputools.deviceId = 0,
    ...
  )

  predictions <- predict(darch, newdata=testData, type="class")
  
  labels <- cbind(predictions, testLabels)
  numIncorrect <- sum(apply(labels, 1, function(i) { any(i[1:10] != i[11:20]) }))
  cat(paste0("Incorrect classifications on test data: ", numIncorrect,
             " (", round(numIncorrect/nrow(testLabels)*100, 2), "%)\n"))
  
  darch
}