I have the following object
Formal class 'dgCMatrix' [package "Matrix"] with 6 slots
..# i : int [1:120671481] 0 2 3 6 10 13 21 22 25 36 ...
..# p : int [1:51366] 0 3024 4536 8694 3302271 3302649 5715381 5756541 5784009 5801691 ...
..# Dim : int [1:2] 10314738 51365
..# Dimnames:List of 2
.. ..$ : chr [1:10314738] "line1" "line2" "line3" "line4" ...
.. ..$ : chr [1:51365] "sparito" "davide," "15enne" "di" ...
.. .. ..- attr(*, ".match.hash")=Class 'match.hash' <externalptr>
..# x : num [1:120671481] 1 1 1 1 1 1 1 1 1 1 ...
..# factors : list()
This object comes from the function dtm_builder of text2map package. Since I would like to remove empty rows from the matrix, I thought about using the command:
raw.sum=apply(dtm,1,FUN=sum) #sum by raw each raw of the table
dtm2=dtm[raw.sum!=0,]
Anyway, I obtained the following error:
Error in asMethod(object): Cholmod error 'problem too large' at file ..
How could I fix it?
The short answer to your problem is that you're likely converting a sparse object to a dense object. Matrix package sparse matrix classes are very memory efficient when a matrix has a lot of zeros (like a DTM) by simply not allocating memory for the zeros.
#akrun's answer should work, but there is a rowSums function in base R and a rowSums function from the Matrix package. You would need to load the Matrix package first.
Here is an example dgCMatrix (note not loading Matrix package yet)
m1 <- Matrix::Matrix(1:9, 3, 3, sparse = TRUE)
m1[1, 1:3] <- 0
class(m1)
If we use the base R rowSums you get the error:
rowSums(m1)
Error in rowSums(dtm): 'x' must be an array of at least two dimensions
If the Matrix package is loaded,rowSums will be replaced with the Matrix package's own method, which works with dgCMatrix. This is also true for the bracket operators [. If you update text2map to version 0.1.5, Matrix is loaded by default.
That is a massive DTM, so you may still run into memory issues -- which will depend on your machine. One thing to note is that removing sparse rows/columns will not help much. So, although words that occur once or twice will make up about 60% of your columns, you will reduce the size in terms of memory more by removing the most frequent words (i.e. words with a number in every row).
Related
I have n matrices of which I am trying to apply nearPD()from the Matrixpackage.
I have done this using the following code:
A<-lapply(b, nearPD)
where b is the list of n matrices.
I now would like to convert the list A into matrices. For an individual matrix I would use the following code:
A<-matrix(runif(n*n),ncol = n)
PD_mat_A<-nearPD(A)
B<-as.matrix(PD_mat_A$mat)
But I am trying to do this with a list. I have tried the following code but it doesn't seem to work:
d<-lapply(c, as.matrix($mat))
Any help would be appreciated. Thank you.
Here is a code so you can try and reproduce this:
n<-10
generate<-function (n){
matrix(runif(10*10),ncol = 10)
}
b<-lapply(1:n, generate)
Here is the simplest method using as.matrix as noted by #nicola in the comments below and (a version using apply) by #cimentadaj in the comments above:
d <- lapply(A, function(i) as.matrix(i$mat))
My original answer, exploiting the nearPD data structure was
With a little fiddling with the nearPD object type, here is an extraction method:
d <- lapply(A, function(i) matrix(i$mat#x, ncol=i$mat#Dim[2]))
Below is some commentary on how I arrived at my answer.
This object is fairly complicated as str(A[[1]]) returns
List of 7
$ mat :Formal class 'dpoMatrix' [package "Matrix"] with 5 slots
.. ..# x : num [1:100] 0.652 0.477 0.447 0.464 0.568 ...
.. ..# Dim : int [1:2] 10 10
.. ..# Dimnames:List of 2
.. .. ..$ : NULL
.. .. ..$ : NULL
.. ..# uplo : chr "U"
.. ..# factors : list()
$ eigenvalues: num [1:10] 4.817 0.858 0.603 0.214 0.15 ...
$ corr : logi FALSE
$ normF : num 1.63
$ iterations : num 2
$ rel.tol : num 0
$ converged : logi TRUE
- attr(*, "class")= chr "nearPD"
You are interested in the "mat" which is accessed by $mat. The # symbols show that "mat" is an s4 object and its components are accessed using #. The components of interest are "x", the matrix content, and "Dim" the dimension of the matrix. The code above puts this information together to extract the matrices from the list of "nearPD" objects.
Below is a brief explanation of why as.matrix works in this case. Note the matrix inside a nearPD object is not a matrix:
is.matrix(A[[1]]$mat)
[1] FALSE
However, it is a "Matrix":
class(A[[1]]$mat)
[1] "dpoMatrix"
attr(,"package")
[1] "Matrix"
From the note in the help file, help("as.matrix,Matrix-method"),
Loading the Matrix namespace “overloads” as.matrix and as.array in the base namespace by the equivalent of function(x) as(x, "matrix"). Consequently, as.matrix(m) or as.array(m) will properly work when m inherits from the "Matrix" class.
So, the Matrix package is taking care of the as.matrix conversion "under the hood."
I know this has been asked before and the answer has always been to use sink but I'm getting the prompt written out too and I don't want that. Any way to just store the output to a .txt file and not the prompts?
info(hdr) prints the following to console:
> info(hdr)
DataFrame with 3 rows and 3 columns
Number Type Description
<character> <character> <character>
NS 1 Integer Number of Samples With Data
DP 1 Integer Total Depth
DB 0 Flag dbSNP membership, build 131
I want to send all of info()'s output to a .txt file and these are the commands I'm using
sink("info.txt")
info(hdr)
sink()
What's written in info.txt:
> info(hdr)
DataFrame with 3 rows and 3 columns
Number Type Description
<character> <character> <character>
NS 1 Integer Number of Samples With Data
DP 1 Integer Total Depth
DB 0 Flag dbSNP membership, build 131
> sink()
Why are the commands showing up too? Anyway to prevent that?
EDIT: for clarity. The function info() comes from VariantAnnotation package from the following link: http://www.bioconductor.org/help/workflows/variants/. The output of str(hdr) is:
Formal class 'VCFHeader' [package "VariantAnnotation"] with 3 slots
..# reference: chr(0)
..# samples : chr "GS06985-1100-37-ASM"
..# header :Formal class 'SimpleDataFrameList' [package "IRanges"] with 4 slots
.. .. ..# elementType : chr "DataFrame"
.. .. ..# elementMetadata: NULL
.. .. ..# metadata : list()
.. .. ..# listData :List of 4
Many more lines but I truncated.
I want to first calculate a markov transition matrix and then take exponent of it. To achieve the first goal I use the markovchainFit function inside markovchain package and it return me a data.frame , rather than a matrix. So I need to convert it to matrix before I take exponent.
My R code snippet is like
#################################
# Estimate Transition Matrix #
#################################
setwd("G:/Data_backup/GDP_per_Capita")
library("foreign")
library("Hmisc")
mydata <- stata.get("G:/Data_backup/GDP_per_Capita/states.dta")
mydata
library(markovchain)
library(expm)
rgdp_e=mydata[,2:7]
rgdp_o=mydata[,8:13]
createSequenceMatrix(rgdp_e)
rgdp_e_trans<-markovchainFit(data=rgdp_e,,method="bootstrap",nboot=5, name="Bootstrap Mc")
rgdp_e_trans<-as.numeric(unlist(rgdp_e_trans))
rgdp_e_trans<-as.matrix(rgdp_e_trans)
is.matrix(rgdp_e_trans)
rgdp_e_trans %^% 1/5
the rgdp_e_trans is a data frame, and I try to convert it to a numeric matrix. It seems work when I test it using is.matrix command. However, the final line give me an error said
Error in rgdp_e_trans %^% 2 :
(list) object cannot be coerced to type 'double'
After some searching work in stackoverflow, I find this question sharing the similar problem and use rgdp_e_trans<-as.numeric(unlist(rgdp_e_trans)) to coerce the object to be `double', but it seems not work.
Besides, the data.frame rgdp_e_trans contains no factor or characters
The output in the console is like
> rgdp_e=mydata[,2:7]
> rgdp_o=mydata[,8:13]
> createSequenceMatrix(rgdp_e)
Error: not compatible with STRSXP
> rgdp_e_trans<-markovchainFit(data=rgdp_e,,method="bootstrap",nboot=5, name="Bootstrap Mc")
> rgdp_e_trans
$estimate
1 2 3 4 5
1 0.6172840 0.18930041 0.09053498 0.074074074 0.02880658
2 0.1125828 0.59602649 0.28476821 0.006622517 0.00000000
3 0.0000000 0.03846154 0.60256410 0.358974359 0.00000000
4 0.0000000 0.01162791 0.03488372 0.691860465 0.26162791
5 0.0000000 0.00000000 0.00000000 0.044247788 0.95575221
> rgdp_e_trans<-as.numeric(unlist(rgdp_e_trans))
Error: (list) object cannot be coerced to type 'double'
> rgdp_e_trans<-as.matrix(rgdp_e_trans)
> is.matrix(rgdp_e_trans)
[1] TRUE
> rgdp_e_trans %^% 1/5
Error in rgdp_e_trans %^% 1 :
(list) object cannot be coerced to type 'double'
>
Any suggestion to fix the problem, or alternative way to calculate the exponent ? Thank you.
Additional：
> str(rgdp_e_trans)
List of 1
$ estimate:Formal class 'markovchain' [package "markovchain"] with 4 slots
.. ..# states : chr [1:5] "1" "2" "3" "4" ...
.. ..# byrow : logi TRUE
.. ..# transitionMatrix: num [1:5, 1:5] 0.617 0.113 0 0 0 ...
.. .. ..- attr(*, "dimnames")=List of 2
.. .. .. ..$ : chr [1:5] "1" "2" "3" "4" ...
.. .. .. ..$ : chr [1:5] "1" "2" "3" "4" ...
.. ..# name : chr "Bootstrap Mc"
and I comment out the as.matrix part
rgdp_e=mydata[,2:7]
rgdp_o=mydata[,8:13]
createSequenceMatrix(rgdp_e)
rgdp_e_trans<-markovchainFit(data=rgdp_e,,method="bootstrap",nboot=5, name="Bootstrap Mc")
rgdp_e_trans
str(rgdp_e_trans)
# rgdp_e_trans<-as.numeric(unlist(rgdp_e_trans))
# rgdp_e_trans<-as.matrix(rgdp_e_trans)
# is.matrix(rgdp_e_trans)
rgdp_e_trans$estimate %^% 1/5
You can access the transition matrix directly from the object returned by markovchainFit as:
rgdp_e_trans$estimate#transitionMatrix
Here rgdp_e_trans is your return value from markovchainFit, which is actually a list containing the information from the fitting process. You access the estimates item of that list by using the $ operator. The estimate object is from a formal S4 class (see e.g. Advanced R by Hadley Wickham for a description of the object systems used in R), which is why in order to access its items you have to use the # operator instead of the standard $ used for the more common S3 objects.
If you print out the return value of as.matrix(rgdp_e_trans) it should be immediately obvious where your initial approach went wrong. In general it's a good idea to check the structure of an object with the str function - instead of relying on its print method - when you encounter unexpected results or are working with new types of objects.
I can't really create a code example because I'm not quite sure what the problem is and my actual problem is rather involved. That said it seems like kind of a generic problem that maybe somebody's seen before.
Basically I'm constructing 3 different dataframes and rbinding them together, which is all as expected smooth sailing but when I try to write that merged frame back to the DB I get this error:
Error in .External2(C_writetable, x, file, nrow(x), p, rnames, sep, eol, :
unimplemented type 'list' in 'EncodeElement'
I've tried manually coercing them using as.data.frame() before and after the rbinds and the returned object (the same one that fails to write with the above error message) exists in the environment as class data.frame so why does dbWriteTable not seem to have got the memo?
Sorry, I'm connecting to a MySQL DB using RMySQL. The problem I think as I look a little closer and try to explain myself is that the columns of my data frame are themselves lists (of the same length), which sorta makes sense of the error. I'd think (or like to think anyways) that a call to as.data.frame() would take care of that but I guess not?
A portion of my str() since it's long looks like:
.. [list output truncated]
$ stcong :List of 29809
..$ : int 3
..$ : int 8
..$ : int 4
..$ : int 2
I guess I'm wondering if there's an easy way to force this coercion?
Hard to say for sure, since you provided so little concrete information, but this would be one way to convert a list column to an atomic vector column:
> d <- data.frame(x = 1:5)
> d$y <- as.list(letters[1:5])
> str(d)
'data.frame': 5 obs. of 2 variables:
$ x: int 1 2 3 4 5
$ y:List of 5
..$ : chr "a"
..$ : chr "b"
..$ : chr "c"
..$ : chr "d"
..$ : chr "e"
> d$y <- unlist(d$y)
> str(d)
'data.frame': 5 obs. of 2 variables:
$ x: int 1 2 3 4 5
$ y: chr "a" "b" "c" "d" ...
This assumes that each element of your list column is only a length one vector. If any aren't, things will be more complicated, and you'd likely need to rethink your data structure anyhow.
I am using the tm package to compute term-document-matrix for a dataset, I now have to write the term-document-matrix to a file but when I use the write functions in R I am getting a error.
Here is the code which I am using and the error I am getting:
data("crude")
tdm <- TermDocumentMatrix(crude, control = list(weighting = weightTfIdf, stopwords = TRUE))
dtm <- DocumentTermMatrix(crude, control = list(weighting = weightTfIdf, stopwords = TRUE))
and this is the error while I use the write.table command on this data:
Error in cat(list(...), file, sep, fill, labels, append) : argument 1 (type 'list') cannot be handled by 'cat'
I understand that tbm is a object of type Simple Triplet Matrix, but how can I write this to a simple text file.
I think I might be misunderstanding the question, but if all you want to do is export the term document matrix to a file, then how about this:
m <- inspect(tdm)
DF <- as.data.frame(m, stringsAsFactors = FALSE)
write.table(DF)
Is that what you're after mate?
Hope that helps a little,
Tony Breyal
Should the file be "human-readable"? If not, use dump, dput, or save. If so, convert your list into a data.frame.
Edit: You can convert your list into a matrix if each list element is equal length by doing matrix(unlist(list.name), nrow=length(list.name[[1]])) or something like that (or with plyr).
Why aren't you doing your SVM analysis in R (e.g. with kernlab)?
Edit 2: Ok, I looked at your data, and it isn't easy to convert into a matrix because the list elements aren't equal length:
> is.list(tdm)
[1] TRUE
> str(tdm)
List of 7
$ i : int [1:1475] 15 29 151 152 173 205 215 216 227 228 ...
$ j : int [1:1475] 1 1 1 1 1 1 1 1 1 1 ...
$ v : Named num [1:1475] 3.32 4.32 2.32 2 2.32 ...
..- attr(*, "names")= chr [1:1475] "1.50" "16.00" "barrel," "barrel." ...
$ nrow : int 985
$ ncol : int 20
$ dimnames :List of 2
..$ Terms: chr [1:985] "(bpd)" "(bpd)." "(gcc)" "(it) appears to be nearing a crossroads with regard to\nderegulation, both as it pertains to investments and imports," ...
..$ Docs : chr [1:20] "127" "144" "191" "194" ...
$ Weighting: chr [1:2] "term frequency - inverse document frequency" "tf-idf"
- attr(*, "class")= chr [1:2] "TermDocumentMatrix" "simple_triplet_matrix"
In order to convert this to a matrix, you will need to either take elements of this list (e.g. i, j) or else do some other manipulation.
Edit 3: Just to conclude my commentary here: these objects are intended to be used with the inspect function (see the package vignette).
As discussed, in order to use a function like write.table, you will need to convert your list into a matrix, which requires some manipulation of that list such that you have several vectors of equal length. Looking at the structure of these tm objects: this will be very difficult to do, and I suggest you work with the helper functions that are included with that package.
dtmMatrix <- as.matrix(dtm)
write.csv(dtmMatrix, 'mydata.csv')
This certainly does the work. However, when I tried it on a very large DTM (25000 by 35000), it gave errors relating to lack of memory space.
I used the following method:
dtm <- DocumentTermMatrix(corpus)
dtm1 <- removeSparseTerms(dtm,0.998) ##max allowed sparsity 0.998
m <- inspect(dtm1)
DF <- as.data.frame(m, stringsAsFactors = FALSE)
write.csv(DF,"mydata0.998sparse.csv")
Which reduced the size of the document term matrix to a great extent!
Here you can increase the max allowable sparsity (closer to 1) to include more terms in DF.