Question

我有一个像这样的矩阵

library(matrixStats)  ## for `rowMaxs`
mymat <- matrix(c(1:10,1:10), 5,4)

#     [,1] [,2] [,3] [,4]
#[1,]    1    6    1    6
#[2,]    2    7    2    7
#[3,]    3    8    3    8
#[4,]    4    9    4    9
#[5,]    5   10    5   10

myvector <- rowMaxs(mymat)

我希望每行中的所有元素都替换为0，期望该行中的最大/最大值。在此示例中，每行有两个最大值，因此预期结果为：

#     [,1] [,2] [,3] [,4]
#[1,]    0    6    0    6
#[2,]    0    7    0    7
#[3,]    0    8    0    8
#[4,]    0    9    0    9
#[5,]    0   10    0   10

Answer 1

## replace all elements to 0 except for row maximum; multiple maximum allowed
## recycling rule is used here, so `myvector` is replicated to be a full matrix
## i.e., equivalently: mymat[ mymat < rep.int(myvector, ncol(mymat)) ] <- 0
mymat[mymat < myvector] <- 0

如果要反向操作，即将行最大值（最大值）设置为0，请使用

mymat[mymat == myvector] <- 0

您已获得myvector matrixStats::rowMaxs。对于想要使用R base的其他人来说，max.col很优雅：

## matrix indexing is used, i.e., we index matrix with a two-column matrix
myvector <- mymat[ cbind( 1:dim(mymat)[1], max.col(mymat) ) ]

Answer 2

我认为我的第一个解决方案将与哲源的帖子一致，但方式略有不同：

mymat[mymat != matrix(rep(rowMaxs(mymat), ncol(mymat)),ncol=ncol(mymat))] <- 0

或申请：

t(apply(mymat, 1, function(x) {x[x != max(x)] <- 0; x}))

两个输出

     [,1] [,2] [,3] [,4]
[1,]    0    6    0    6
[2,]    0    7    0    7
[3,]    0    8    0    8
[4,]    0    9    0    9
[5,]    0   10    0   10

矩阵操作：将所有元素设置为0，但行最大值（最大值）除外

2 个答案: