aperm clarification in R - r

Question

I am following the thread 2d matrix to 3d stacked array in r and have a clarification on the aperm function.
1) I get the first part of the solution, but did not understand the c(2,1,3) used in the function. Could you kindly clarify that?
2) Also I am trying a slight variation of the example in that thread.
My case is as follows:
For a similar matrix in example:
set.seed(1)
mat <- matrix(sample(100, 12 * 5, TRUE), ncol = 5)
[,1] [,2] [,3] [,4] [,5]
[1,] 27 69 27 80 74
[2,] 38 39 39 11 70
[3,] 58 77 2 73 48
[4,] 91 50 39 42 87
[5,] 21 72 87 83 44
[6,] 90 100 35 65 25
[7,] 95 39 49 79 8
[8,] 67 78 60 56 10
[9,] 63 94 50 53 32
[10,] 7 22 19 79 52
[11,] 21 66 83 3 67
[12,] 18 13 67 48 41
I am trying to rearrange such that I have a 3 (row) X 5 (col) x 11 (third dim) array.
So, essentially the rows would overlap and show something like:
,,1
27 69 27 80 74
38 39 39 11 70
58 77 2 73 48
,,2
38 39 39 11 70
58 77 2 73 48
91 50 39 42 87
,,3
58 77 2 73 48
91 50 39 42 87
21 72 87 83 44
and so on until we hit ,,11
Would someone have any experience with this?
Thanks!

Answer 1

Just stumbled over this question. Though the answer comes a little late, here are two options for you.
First, you need to extend mat in such a way that it's rows overlap. We can use this vector for row indexing.
#[1] 1 2 3 2 3 4 3 4 5 4 5 6 5 6 7 6 7 8 7 8 9 8 9 10 9 10 11 10 11 12
I used rollapply from the zoo package to create it as follows:
library(zoo)
row_nums <- c(t(rollapply(1:nrow(mat), width = 3, FUN = rep, 1)))
mat <- mat[row_nums, ]
dim(mat)
#[1] 30 5
Now use the matsplitter function that #Mr.Flick provided in this answer (please consider to upvote his answer) to get the desired output:
matsplitter(mat, 3, 5)
#, , 1
#
# [,1] [,2] [,3] [,4] [,5]
#[1,] 27 69 27 80 74
#[2,] 38 39 39 11 70
#[3,] 58 77 2 73 48
#
#, , 2
#
# [,1] [,2] [,3] [,4] [,5]
#[1,] 38 39 39 11 70
#[2,] 58 77 2 73 48
#[3,] 91 50 39 42 87
#
#, , 3
#
# [,1] [,2] [,3] [,4] [,5]
#[1,] 58 77 2 73 48
#[2,] 91 50 39 42 87
#[3,] 21 72 87 83 44
#
#, , 4
# ...
Note that you will end up with an array of dimension 3 x 5 x 10, not 11.
matsplitter <- function(M, r, c) {
rg <- (row(M) - 1) %/% r + 1
cg <- (col(M) - 1) %/% c + 1
rci <- (rg - 1) * max(cg) + cg
N <- prod(dim(M)) / r / c
cv <- unlist(lapply(1:N, function(x)
M[rci == x]))
dim(cv) <- c(r, c, N)
cv
}
Here is a solution using aperm as in the linked answer (assuming that mat was extended as above and is of dimension 30 x 5).
aperm(`dim<-`(t(mat), list(5, 3, 10)), c(2, 1, 3))
t(mat): transposes mat (new dimension: 5 x 30)
`dim<-`(t(mat), list(5, 3, 10)): changes the dimension of t(mat) from 5 X 30 to 5 x 3 x 10
aperm(..., c(2, 1, 3)) permutes the dimensions of the array `dim<-`(t(mat), list(5, 3, 10)) from 5 x 3 x 10 to 3 x 5 x 10, i.e. the second dimension becomes the first, the first
dimension becomes the second and the third dimension stays the same.