SparseM.hb {SparseM}R Documentation

Harwell-Boeing Format Sparse Matrices

Description

Read, and extract components of data in Harwell-Boeing sparse matrix format.

Usage

read.matrix.hb(file)
model.matrix(object, ...)
model.response(data,type)

Arguments

file

file name to read from or

data, object

an object of either 'matrix.csc.hb' or 'matrix.ssc.hb' class

type

One of ‘"any"’, ‘"numeric"’, ‘"double"’. Using the either of latter two coerces the result to have storage mode ‘"double"’

...

additional arguments to model.matrix

Details

Sparse coefficient matrices in the Harwell-Boeing format are stored in 80-column records. Each file begins with a multiple line header block followed by two, three or four data blocks. The header block contains summary information on the storage formats and storage requirements. The data blocks contain information of the sparse coefficient matrix and data for the right-hand-side of the linear system of equations, initial guess of the solution and the exact solutions if they exist. The function model.matrix extracts the X matrix component. The function model.response extracts the y vector (or matrix). The function model.guess extracts the guess vector. The function model.xexact extracts the xexact vector. This function is written in R replacing a prior implementation based on iohb.c which had memory fault difficulties. The function write.matrix.hb has been purged; users wishing to write matrices in Harwell-Boeing format are advised to convert SparseM matrices to Matrix classes and use writeHB from the Matrix package. Contributions of code to facilitate this conversion would be appreciated!

Value

The function read.matrix.hb returns a list of class matrix.csc.hb or matrix.ssc.hb depending on how the coefficient matrix is stored in the file.

ra

ra component of the csc or ssc format of the coefficient matrix, X.

ja

ja component of the csc or ssc format of the coefficient matrix, X.

ia

ia component of the csc or ssc format of the coefficient matrix, X.

rhs.ra

ra component of the right-hand-side, y, if stored in csc or ssc format; right-hand-side stored in dense vector or matrix otherwise.

rhs.ja

ja component of the right-hand-side, y, if stored in csc or ssc format; a null vector otherwise.

rhs.ia

ia component of the right-hand-side, y, if stored in csc or ssc format; a null vector otherwise.

xexact

vector of the exact solutions, b, if they exist; a null vector otherwise.

guess

vector of the initial guess of the solutions if they exist; a null vector otherwise.

dimension

dimenson of the coefficient matrix, X.

rhs.dim

dimenson of the right-hand-side, y.

rhs.mode

storage mode of the right-hand-side; can be full storage or same format as the coefficient matrix, for the moment the only allowed mode is "F" for full, or dense mode.

The function model.matrix returns the X matrix of class matrix.csr. The function model.response returns the y vector (or matrix). The function model.guess returns the guess vector (or matrix). The function model.xexact returns the xexact vector (or matrix).

Author(s)

Pin Ng

References

Duff, I.S., Grimes, R.G. and Lewis, J.G. (1992) User's Guide for Harwell-Boeing Sparse Matrix Collection at http://math.nist.gov/MatrixMarket/collections/hb.html

See Also

slm for sparse version of lm
SparseM.ops for operators on class matrix.csr
SparseM.solve for linear equation solving for class matrix.csr
SparseM.image for image plotting of class matrix.csr
SparseM.ontology for coercion of class matrix.csr

Examples

 
Xy <- read.matrix.hb(system.file("extdata","lsq.rra",package = "SparseM"))
class(Xy) # -> [1] "matrix.csc.hb"
X <- model.matrix(Xy)->X
class(X) # -> "matrix.csr"
dim(X) # -> [1] 1850  712
y <- model.response(Xy) # extract the rhs
length(y) # [1] 1850
Xy <- read.matrix.hb(system.file("extdata","rua_32_ax.rua",package = "SparseM"))
X <- model.matrix(Xy)
y <- model.response(Xy) # extract the rhs
g <- model.guess(Xy) # extract the guess 
a <- model.xexact(Xy) # extract the xexact 
fit <- solve(t(X) %*% X, t(X) %*% y) # compare solution with xexact solution 

[Package SparseM version 1.77 Index]