Question

我有两个矩阵。我想使用第一列的列来过滤第二列，然后找到过滤集的总和。我使用了以下代码，它完全正常。

apply(firstMat,2,function(x) sum(secondMat[x,x]))

但是，数据集很大，我想找到一种替代方法，使流程更快。

以下是小规模的可重复示例：

firstMat<-matrix(c(T,F,T,F,F,T,T,F,F,F),nrow=5,ncol=2)
secondMat<-matrix(c(1,0,0,0,1,0,0,0,1,1,1,0,1,0,1,1,1,0,0,0,1,1,1,0,1),nrow=5,ncol=5)

如果你能帮助我，我将非常感激。

Answer 1

您可以在多个群集上并行运行apply功能

firstMat<-matrix(c(T,F,T,F,F,T,T,F,F,F),nrow=5,ncol=2)
secondMat<-matrix(c(1,0,0,0,1,0,0,0,1,1,1,0,1,0,1,1,1,0,0,0,1,1,1,0,1),nrow=5,ncol=5)

# create custers
library(doSNOW)
cl <- makeCluster(2, type = "SOCK") # creates 2 clusters 
# can use detectCores() from package parallel to check number of cores in your machine
registerDoSNOW(cl)
clusterExport(cl,list("secondMat")) # need to export secndMAT to each cluster since will be used in cluster

# Option 1: Using parApply from package `parallel`
library(parallel)
parApply(cl,firstMat,2,function(x) sum(secondMat[x,x]))

# Option 2: Using aaply from package `plyr`
library(plyr)    
aaply(firstMat,2,function(x) sum(secondMat[x,x]),.parallel=T)

stopCluster(cl)

使用可重复性较小的示例，它没有显示任何速度改进，但我认为对于大型矩阵，两个选项都比apply更快

Answer 2

也许你的BLAS比显式循环更快：

diag( t(firstMat) %*% secondMat %*% firstMat )

如何使apply（）函数更快？

2 个答案: