r - Parallel allocation of a matrix

Asked Mar 10 '17 at 14:28

Active Mar 10 '17 at 14:54

Viewed 24 times

I have a vector, v. I am trying to construct a matrix whose elements M[i][j] depend on v[i] and v[j].

The 'loop' way of performing this is the following (for example):

v = c(1:4)
M = matrix(0, nrow = length(v), ncol = length(v))    
for(i in 1:length(v)){
    for(j in i:length(v)){
        M[i,j] = v[i] * v[j]
    }
}

M then looks like:

     [,1] [,2] [,3] [,4]
[1,]    1    2    3    4
[2,]    0    4    6    8
[3,]    0    0    9   12
[4,]    0    0    0   16

Where the matrix can now be easily symmetrised.

However, this become prohibitively slow when v becomes large. Considering how embarrassingly parallel this problem is, I was hoping there was a way of performing this task in parallel?

NB - The actual function I am performing on the elements of the vector is more complicated than multiplication, but that's just a decent example!

edited Mar 10 '17 at 14:54

asked Mar 10 '17 at 14:28

Lee Steinberg

Can you make your example reproducible with what the desired output should look like? – Roman Luštrik Mar 10 '17 at 14:30
Assuming you function `f` is a simple sum, wouldn't the result be `outer(v, v, '+')` – talat Mar 10 '17 at 14:34
I have edited the question to give more information, hope thats useful! The issue with `outer` is that it gets prohibitively slow on large vectors. – Lee Steinberg Mar 10 '17 at 14:39

r - Parallel allocation of a matrix

0 Answers0