Re: [R] Possible bug of QR decomposition in package Matrix
Thank you, Martin. Yes, on my computer (48Gb memory) it also gives a message "caught segfault" as you got. So it is because the matrix size is too large. I apologize for the offence I've caused. 2011/8/4 Martin Maechler : >>>>>> C6H5NO2 >>>>>> on Thu, 4 Aug 2011 16:03:54 +0800 writes: > > > Thank you very much, Josh! > > As you suggested, I will contact the developers of "Matrix". > > > PS, C6 are just initial characters of my email account :-) > > > Best wishes, > > C6 > > well, as the posting guide (http://www.R-project.org/posting-guide.html) > says, this is regarded as impolite by many, > and if I wasn't one of the Matrix package authors, > I would not spend time helping 'C6' either. > > > 2011/8/4 Joshua Wiley : > >> Hi C6 (were C1 - 5 already taken in your family?), > >> > >> I downloaded your data and can replicate your problem. R > >> ceases responding and terminates. This does not occur with all > >> uses of qr on a dgCMatrix object. I know nothing about sparse > >> matrices, but if you believe this should not be occurring, you > >> should contact the package maintainers. Here is my > >> sessionInfo() (FYI, it would probably be helpful to report > >> yours also in case the issue is version dependent): > >> > >> R Under development (unstable) (2011-07-30 r56564) Platform: > >> x86_64-pc-mingw32/x64 (64-bit) > >> > >> locale: [1] LC_COLLATE=English_United States.1252 [2] > >> LC_CTYPE=English_United States.1252 [3] > >> LC_MONETARY=English_United States.1252 [4] 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] Matrix_0.999375-50 lattice_0.19-30 > >> > >> loaded via a namespace (and not attached): [1] grid_2.14.0 > >> tools_2.14.0 > >> > >> Cheers, > >> > >> Josh > >> > >> On Wed, Aug 3, 2011 at 4:26 PM, C6H5NO2 > >> wrote: > >>> Hello R users, > >>> > >>> I am trying to give the QR decomposition for a large sparse > >>> matrix in the format of dgCMatrix. When I run qr function for > >>> this matrix, the R session simply stops and exits to the > >>> shell. The matrix is of size 108595x108595, and it has > >>> 4866885 non-zeros. I did the experiment on windows 7 and linux > >>> mint 11 (both 64 bit), and the results are the same. > >>> > >>> I have uploaded my data file to > >>> http://ifile.it/elf2p6z/A.RData . The file is 10.681 MB and I > >>> hope someone could kindly download it. The code to see my > >>> problem is: > > >>> library(Matrix) > >>> load("A.RData") > >>> B <- qr(A) > > > >>> Best wishes, C6 > > And what's the size of RAM your two computers have ?? > The answer is of quite some importance. > > Short answer: If you have a large very sparse matrix, > you don't know if the QR decomposition of that matrix is also > very sparse... and if it ain't it will blow up memory, > and that's what I'm pretty sure happened with you. > > What I don't see is why R "simply stops" for you and does not > through a an error message about insufficient memory. > As I show below, I do get a seg.fault > --- which may be considered a bug --- > *BUT* I do get the message about memory problems. > Did you really *not* get any such message? > Is it because you've used a GUI that hides such valuable > information from the user? > > Here's the more detailed reason / analysis about why the above > "kills R". This is commented R code, > you can cut paste after you've got 'A' : > > str(A) > ## Formal class 'dgCMatrix' [package "Matrix"] with 6 slots > ## ..@ i : int [1:4866885] 0 1 2 16 32 33 2392 2417 0 1 ... > ## ..@ p : int [1:108596] 0 8 21 35 44 51 59 63 69 78 ... > ## ..@ Dim : int [1:2] 108595 108595 > ## ..@ Dimnames:List of 2 > ## .. ..$ : NULL > ## .. ..$ : NULL > ## ..@ x : num [1:4866885] 140.03 14.79 14.79 1.78 1.78 ... > ## ..@ factors : list() > > system.time(# the following is still not as fast a
Re: [R] Possible bug of QR decomposition in package Matrix
Thank you very much, Josh! As you suggested, I will contact the developers of "Matrix". PS, C6 are just initial characters of my email account :-) Best wishes, C6 2011/8/4 Joshua Wiley : > Hi C6 (were C1 - 5 already taken in your family?), > > I downloaded your data and can replicate your problem. R ceases > responding and terminates. This does not occur with all uses of qr on > a dgCMatrix object. I know nothing about sparse matrices, but if you > believe this should not be occurring, you should contact the package > maintainers. Here is my sessionInfo() (FYI, it would probably be > helpful to report yours also in case the issue is version dependent): > > R Under development (unstable) (2011-07-30 r56564) > Platform: x86_64-pc-mingw32/x64 (64-bit) > > locale: > [1] LC_COLLATE=English_United States.1252 > [2] LC_CTYPE=English_United States.1252 > [3] LC_MONETARY=English_United States.1252 > [4] 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] Matrix_0.999375-50 lattice_0.19-30 > > loaded via a namespace (and not attached): > [1] grid_2.14.0 tools_2.14.0 > > Cheers, > > Josh > > On Wed, Aug 3, 2011 at 4:26 PM, C6H5NO2 wrote: >> Hello R users, >> >> I am trying to give the QR decomposition for a large sparse matrix in >> the format of dgCMatrix. When I run qr function for this matrix, the R >> session simply stops and exits to the shell. >> The matrix is of size 108595x108595, and it has 4866885 non-zeros. I >> did the experiment on windows 7 and linux mint 11 (both 64 bit), and >> the results are the same. >> >> I have uploaded my data file to http://ifile.it/elf2p6z/A.RData . The >> file is 10.681 MB and I hope someone could kindly download it. >> The code to see my problem is: >> library(Matrix) >> load("A.RData") >> B <- qr(A) >> >> Best wishes, >> C6 >> >> __ >> 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. >> > > > > -- > Joshua Wiley > Ph.D. Student, Health Psychology > Programmer Analyst II, ATS Statistical Consulting Group > University of California, Los Angeles > https://joshuawiley.com/ > __ 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.
[R] Possible bug of QR decomposition in package Matrix
Hello R users, I am trying to give the QR decomposition for a large sparse matrix in the format of dgCMatrix. When I run qr function for this matrix, the R session simply stops and exits to the shell. The matrix is of size 108595x108595, and it has 4866885 non-zeros. I did the experiment on windows 7 and linux mint 11 (both 64 bit), and the results are the same. I have uploaded my data file to http://ifile.it/elf2p6z/A.RData . The file is 10.681 MB and I hope someone could kindly download it. The code to see my problem is: library(Matrix) load("A.RData") B <- qr(A) Best wishes, C6 __ 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.
[R] columnwise nnzero for dgCMatrix
Hi R users, I want to know whether there is a fast method to compute the nonzeroes for each column of dgCMatrix. For summation or average I can use colSums or colMeans. To count the non-zeroes I write a function colCounts <- function(Mat) { M1 <- apply(Mat, MARGIN=2, FUN=nnzero) } But it seems this is quite slow because the apply function would transform a dgCMatrix to a matrix. So I come here to ask if there is anybody know some fast method. -- View this message in context: http://r.789695.n4.nabble.com/columnwise-nnzero-for-dgCMatrix-tp3625726p3625726.html Sent from the R help mailing list archive at Nabble.com. __ 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.