Skip to content

Row-wise kronecker product with Matrix package

2 messages · Ally, ilai

Original
Ally
# 
I'm trying to calculate the row-wise kronecker product A \Box B of two
sparse matrices A and B, and am struggling to find a quick way to do this
that takes advantage of sparseness.  I thought a good idea would be to use
"rep" to construct 2 matrices of the same dimension of the end product, and
multiply these two together: 

library(Matrix)
A<-Matrix(c(1,0,0,0,0,1,2,0), 2, 4)
B<-Matrix(c(2,5,0,0,0,1,0,0,0,0), 2, 5)

A[, rep(seq(ncol(A)), each = ncol(B))] * B[, rep(seq(ncol(B)),ncol(A))]
		
This works, but for much larger problems is slow (compared to keeping A and
B dense).  I was wondering why this happens, and whether there might be a
way around it?  

Thanks in advance for any advice, 

Alastair

--
View this message in context: http://r.789695.n4.nabble.com/Row-wise-kronecker-product-with-Matrix-package-tp4373437p4373437.html
Sent from the R help mailing list archive at Nabble.com.
ilai
# 
Maybe one of these will improve:
...
spam::kronecker         Kronecker Products on Sparse Matrices
spam::spam.class        Class "spam"
base::kronecker         Kronecker Products on Arrays
Matrix::kronecker-methods
                        Methods for Function 'kronecker()' in Package
                        'Matrix'

I doubt it because they will calculate the "full" Kronecker prod and
it will be up to you to index the rows, but you never know...

 system.time(A[, rep(seq(ncol(A)), each = ncol(B))] *
B[,rep(seq(ncol(B)),ncol(A))])
   user  system elapsed
  0.016   0.000   0.019
user  system elapsed
  0.008   0.000   0.008
user  system elapsed
  0.008   0.000   0.009

Cheers
On Thu, Feb 9, 2012 at 9:38 AM, Ally <a.rushworth at stats.gla.ac.uk> wrote: