我有一个稀疏矩阵列表,它们具有相同的行数但列数不同。
这是一个玩具数据集:
library(dplyr)
library(Matrix)
ms <- list(
m1 = data.frame(a = c(1, 10, 100), d = c(2, 20, 200), e = c(3, 30, 300)) %>% as.matrix %>% as("sparseMatrix"),
m2 = data.frame(a = c(4, 40, 400), e = c(5, 50, 500), f = c(6, 60, 600), g = c(7, 70, 700)) %>% as.matrix%>% as("sparseMatrix"),
m3 = data.frame(c = c(8, 80, 800), d = c(9, 90, 900)) %>% as.matrix%>% as("sparseMatrix")
)
我想按列添加ms
中的每个矩阵。这就是我目前的做法:
# get a list of unique columns
final_names <- sapply(ms, colnames) %>% unlist %>% unique
# create an empty sparseMatrix of those dimensions
final_matrix <- matrix(0, nrow = nrow(ms$m1), ncol = length(final_names)) %>%
set_colnames(final_names) %>% as("sparseMatrix")
# add the matrices by column
for(mat in ms) {
current_colnames <- colnames(mat)
final_matrix[, current_colnames] <- mat + final_matrix[, current_colnames]
}
这是我的输出:
final_matrix
3 x 6 sparse Matrix of class "dgCMatrix"
a d e f g c
[1,] 5 11 8 6 7 8
[2,] 50 110 80 60 70 80
[3,] 500 1100 800 600 700 800
这是有效的,但是当我在真实数据集上尝试它时,我得到了一个分段错误,因此必须有一种更好的方法来创建一个空的稀疏矩阵或其他方法。有什么想法吗?
答案 0 :(得分:1)
NM = unique(unlist(lapply(ms, colnames)))
temp = do.call(cbind, ms)
sapply(NM, function(nm) rowSums(as.matrix(temp[,colnames(temp) %in% nm])))
# a d e f g c
#[1,] 5 11 8 6 7 8
#[2,] 50 110 80 60 70 80
#[3,] 500 1100 800 600 700 800
OR
temp = do.call(cbind, lapply(ms, function(x) as.data.frame(as.matrix(x))))
sapply(split.default(temp, unlist(sapply(ms, colnames))), rowSums)
# a c d e f g
#[1,] 5 8 11 8 6 7
#[2,] 50 80 110 80 60 70
#[3,] 500 800 1100 800 600 700