Convert a matrix to a 1 dimensional array

Question

I have a matrix (32X48).

How can I convert the matrix into a single dimensional array?

Answer 1

you can use as.vector(). It looks like it is the fastest method according to my little benchmark, as follows:

library(microbenchmark)
x=matrix(runif(1e4),100,100) # generate a 100x100 matrix
microbenchmark(y<-as.vector(x),y<-x[1:length(x)],y<-array(x),y<-c(x),times=1e4)

The first solution uses as.vector(), the second uses the fact that a matrix is stored as a contiguous array in memory and length(m) gives the number of elements in a matrix m. The third instantiates an array from x, and the fourth uses the concatenate function c(). I also tried unmatrix from gdata, but it's too slow to be mentioned here.

Here are some of the numerical results I obtained:

> microbenchmark(
        y<-as.vector(x),
        y<-x[1:length(x)],
        y<-array(x),
        y<-c(x),
        times=1e4)

Unit: microseconds
                expr    min      lq     mean  median      uq       max neval
   y <- as.vector(x)  8.251 13.1640 29.02656 14.4865 15.7900 69933.707 10000
 y <- x[1:length(x)] 59.709 70.8865 97.45981 73.5775 77.0910 75042.933 10000
       y <- array(x)  9.940 15.8895 26.24500 17.2330 18.4705  2106.090 10000
           y <- c(x) 22.406 33.8815 47.74805 40.7300 45.5955  1622.115 10000

Flattening a matrix is a common operation in Machine Learning, where a matrix can represent the parameters to learn but one uses an optimization algorithm from a generic library which expects a vector of parameters. So it is common to transform the matrix (or matrices) into such a vector. It's the case with the standard R function optim().

Answer 2

Simple and fast since a 1d array is essentially a vector

vector <- array[1:length(array)]

Answer 3

If you instead had a data.frame (df) that had multiple columns and you want to vectorize you can do

as.matrix(df, ncol=1)

Answer 4

array(A) or array(t(A)) will give you a 1-d array.