问题
My objective is to create:
- a randomly populated matrix with entries either
0or1. In this particular case, the matrix is4x24. - The row sum of each of the
4rows is exactly6. - The column sum of each of the
24columns is exactly1
Call the desired matrix M.
Another way of looking at M:
- There are exactly
24entries equal to1. - No column has more than one
1entry.
Progress:
There are 6 spots on each row with a 1 entry. The rest are zero, the matrix is sparse. With 4 rows, this means that M can be uniquely determined by a matrix of indices that stores the locations of the 1 entries. Call this matrix of indices indexM.
I populated indexM with the numbers 1:24 sampled without replacement:
set.seed(30592)
colNum <- 24
rowSum <-6
numZeros <- colNum-rowSum
OneRow<-c(rep(1,rowSum),rep(0,numZeros))
indexM<-matrix(sample(1:24,replace=FALSE),
nrow=4,ncol=6,byrow=TRUE)
For the given seed, the matrix is: https://pastebin.com/8T21MiDv .
How do I turn indexM into the desired sparse matrix?
I found the sparseMatrix in the Matrix library, but it wants a vector or row indices and another vector of column indices, which is not what I have.
Thank you.
回答1:
I found the
sparseMatrixin theMatrixlibrary, but it wants a vector or row indices and another vector of column indices, which is not what I have.
The constraints impose that...
- row indices are
rep(1:4, 6) - col indices are
1:24
The match between row and col indices is randomized. We can...
library(Matrix)
# fix rows, jumble cols
sparseMatrix(rep(1:4, each=6), sample(1:24))
# fix cols, jumble rows
sparseMatrix(sample(rep(1:4, each=6)), 1:24)
# jumbl'm all
sparseMatrix(sample(rep(1:4, each=6)), sample(1:24))
any of which will return something like
4 x 24 sparse Matrix of class "ngCMatrix"
[1,] . . . . | | . . | . . . | | . | . . . . . . . .
[2,] . | | . . . | . . | . . . . . . . . | . . . | .
[3,] . . . | . . . | . . | | . . . . | . . . | . . .
[4,] | . . . . . . . . . . . . . | . . | . | . | . |
来源:https://stackoverflow.com/questions/45593877/random-binary-matrix-with-row-and-column-sum-constraints