I have a large function computing an iterative algorithm for fitting mixed
linear models. Almost all code relies on functions from the Matrix package.
I've come across an issue that I do not believe previously occurred in earlier
versions of R or Matrix.
I have a large, sparse matrix, A as
> class(A);dim(A)
[1] "dgCMatrix"
attr(,"package")
[1] "Matrix"
[1] 12312 12312
I am in a position where I must find its inverse. I realize this is less than
ideal, and I have two ways of doing this
A.Inv <- solve(A, Ir) or just solve(A)
Where Ir is an identity matrix with the same dimensions as A and it is also
sparse
> class(Ir)
[1] "ddiMatrix"
attr(,"package")
[1] "Matrix"
The issue, however, is that the inverse of A is converted into a dense matrix
and this becomes a huge memory hog, causing the rest of the algorithm to fail.
In prior versions this remained as a sparse matrix.
> A.Inv[1:5, 1:5]
5 x 5 Matrix of class "dgeMatrix"
[,1] [,2] [,3] [,4] [,5]
[1,] 0.6878713 0.0000000 0.0000000 0.0000000 0.0000000
[2,] 0.0000000 0.6718767 0.0000000 0.0000000 0.0000000
[3,] 0.0000000 0.0000000 0.5076945 0.0000000 0.0000000
[4,] 0.0000000 0.0000000 0.0000000 0.2324122 0.0000000
[5,] 0.0000000 0.0000000 0.0000000 0.0000000 0.2139975
I could coerce this matrix to become sparse such as
> AA <- as(A.Inv, 'sparseMatrix')
> class(AA)
[1] "dgCMatrix"
attr(,"package")
[1] "Matrix"
> AA[1:5, 1:5]
5 x 5 sparse Matrix of class "dgCMatrix"
[1,] 0.6878713 . . . .
[2,] . 0.6718767 . . .
[3,] . . 0.5076945 . .
[4,] . . . 0.2324122 .
[5,] . . . . 0.2139975
But I don't think this is best.
So, my question is why is a matrix that is sparse turning into a dense matrix?
Can I avoid that and keep it sparse without having to coerce it to be sparse
after it is created?
Thank you very much
Harold
> sessionInfo()
R version 3.0.1 (2013-05-16)
Platform: x86_64-w64-mingw32/x64 (64-bit)
locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252
[3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C
[5] LC_TIME=English_United States.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] lme4_0.999999-2 Matrix_1.0-12 lattice_0.20-15
loaded via a namespace (and not attached):
[1] grid_3.0.1 nlme_3.1-109 stats4_3.0.1 tools_3.0.1
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
