MNIST_to_dataframes.r

# based on https://gist.github.com/daviddalpiaz/ae62ae5ccd0bada4b9acd6dbc9008706

# load image files
load_image_file = function(filename) {
  f = gzfile(filename, 'rb')
  readBin(f, 'integer', n = 1, size = 4, endian = 'big')
  n    = readBin(f, 'integer', n = 1, size = 4, endian = 'big')
  nrow = readBin(f, 'integer', n = 1, size = 4, endian = 'big')
  ncol = readBin(f, 'integer', n = 1, size = 4, endian = 'big')
  x = readBin(f, 'integer', n = n * nrow * ncol, size = 1, signed = FALSE)
  close(f)
  data.frame(matrix(x, ncol = nrow * ncol, byrow = TRUE))
}

# load label files
load_label_file = function(filename) {
  f = gzfile(filename, 'rb')
  readBin(f, 'integer', n = 1, size = 4, endian = 'big')
  n = readBin(f, 'integer', n = 1, size = 4, endian = 'big')
  y = readBin(f, 'integer', n = n, size = 1, signed = FALSE)
  close(f)
  y
}

# load training data
train = load_image_file('mnist/train-images-idx3-ubyte.gz')
train$y = as.factor(load_label_file('mnist/train-labels-idx1-ubyte.gz'))
#load testing data
test  = load_image_file('mnist/t10k-images-idx3-ubyte.gz')
test$y  = as.factor(load_label_file('mnist/t10k-labels-idx1-ubyte.gz'))

# Now plot the first 100 images + labels
# This section obtained from 
# https://www.r-bloggers.com/2015/11/a-little-h2o-deeplearning-experiment-on-the-mnist-data-set/
par( mfrow = c(10,10), mai = c(0,0,0,0))
for(i in 1:100){
  y = as.matrix(train[i, 1:784])
  dim(y) = c(28, 28)
  image( y[,nrow(y):1], axes = FALSE, col = gray(255:0 / 255))
  text( 0.2, 0, train[i,785], cex = 3, col = 2, pos = c(3,4))
}