How to select 100 examples of each mnist digit in R?

Question

I need to select 100 examples of each digit from the mnist data base (so 100 zeros, 100 ones, 100 twos, etc), how do I do that? I've tried but I cannot find a way.

You might try this [How to read MNIST database in R?](https://stackoverflow.com/questions/21521571/how-to-read-mnist-database-in-r). — dcarlson, Apr 28 '22 at 20:46
`library(keras) mnist <- dataset_mnist() x_train <- mnist$train$x y_train <- mnist$train$y d <- x_train[sample(y_train == 0, 100, replace = FALSE)]` — Tainara Zandoná, Apr 29 '22 at 18:07
@dcarlson I tried the code above but it gives me a large integer instead of the 100 samples of images with label = 0 — Tainara Zandoná, Apr 29 '22 at 18:13

score 0 · Answer 1 · answered Apr 28 '22 at 21:02

library("keras")
library("dplyr")

#choosing just training data
mnist<-dataset_mnist()
train_images = mnist$train$x
train_labels = mnist$train$y

#prepear list for labels and 1000*28*28 array for images
mnist_labels=c()
mnist_images<-array(0,dim=c(1000,28,28))

i=1
for (x in 0:9){
  #find first 100 x labels and save indexes in tempList, later append mnist_label list, mnist_images array with it
  tempList=which(train_labels %in% x)%>%head(100)
  for (y in tempList){
    mnist_labels[i]<-train_labels[y]
    mnist_images[i,1:28,1:28]<-train_images[y,1:28,1:28]
    i=i+1
  }
}
rm(x,y,i)

#Visualize 1th digit
first_digit<- mnist_images[1, 1:28, 1:28]
par(pty="s")
image(t(first_digit), col = gray((0:255)/255), axes= FALSE)

# visualize first 100 digits ( all zeros) if you want to get eg last 100 ( all nines, change under for loop both index with index+900)
par(mfcol=c(10,10))
par(mar=c(0, 1, 1, 0), xaxs='i', yaxs='i')
for (index in 1:100) { 
  im <- mnist_images[index,,]
  im <- t(apply(im, 2, rev)) 
  image(1:28, 1:28, im, col=gray((0:255)/255), 
        xaxt='n', main=paste(mnist_labels[index]),axes= FALSE)
}
rm(index)

How to select 100 examples of each mnist digit in R?

1 Answers1