Skip to content

colRowMedians

Jump to bottom Edit New page
hb edited this page Mar 3, 2015 · 6 revisions

matrixStats: Benchmark report

colMedians() and rowMedians() benchmarks

This report benchmark the performance of colMedians() and rowMedians() against alternative methods.

Alternative methods

  • apply() + median()

Data type "integer"

Data

> rmatrix <- function(nrow, ncol, mode = c("logical", "double", "integer", "index"), range = c(-100, 
+     +100), naProb = 0) {
+     mode <- match.arg(mode)
+     n <- nrow * ncol
+     if (mode == "logical") {
+         X <- sample(c(FALSE, TRUE), size = n, replace = TRUE)
+     }     else if (mode == "index") {
+         X <- seq_len(n)
+         mode <- "integer"
+     }     else {
+         X <- runif(n, min = range[1], max = range[2])
+     }
+     storage.mode(X) <- mode
+     if (naProb > 0) 
+         X[sample(n, size = naProb * n)] <- NA
+     dim(X) <- c(nrow, ncol)
+     X
+ }
> rmatrices <- function(scale = 10, seed = 1, ...) {
+     set.seed(seed)
+     data <- list()
+     data[[1]] <- rmatrix(nrow = scale * 1, ncol = scale * 1, ...)
+     data[[2]] <- rmatrix(nrow = scale * 10, ncol = scale * 10, ...)
+     data[[3]] <- rmatrix(nrow = scale * 100, ncol = scale * 1, ...)
+     data[[4]] <- t(data[[3]])
+     data[[5]] <- rmatrix(nrow = scale * 10, ncol = scale * 100, ...)
+     data[[6]] <- t(data[[5]])
+     names(data) <- sapply(data, FUN = function(x) paste(dim(x), collapse = "x"))
+     data
+ }
> data <- rmatrices(mode = mode)

Results

10x10 integer matrix

> X <- data[["10x10"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   656647 35.1    1168576  62.5  1168576  62.5
Vcells 12135116 92.6   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654509 35.0    1168576  62.5  1168576  62.5
Vcells 12128708 92.6   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on integer+10x10 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 0.0081 0.0154 0.0370 0.0212 0.0447 0.8034
apply+median 0.4823 0.5386 0.7707 0.7634 0.8240 3.2590
expr min lq mean median uq max
colMedians 1.00 1.00 1.00 1.00 1.00 1.000
apply+median 59.66 34.97 20.81 36.05 18.45 4.056
Table: Benchmarking of rowMedians() and apply+median() on integer+10x10 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 0.0085 0.0108 0.0222 0.0183 0.0264 0.0708
apply+median 0.4862 0.5018 0.6106 0.5216 0.7383 1.1772
expr min lq mean median uq max
rowMedians 1.0 1.00 1.00 1.00 1 1.00
apply+median 57.4 46.55 27.48 28.52 28 16.62
Figure: Benchmarking of colMedians() and apply+median() on integer+10x10 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on integer+10x10 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
2 rowMedians 8.470 10.78 22.22 18.29 26.37 70.83
1 colMedians 8.085 15.40 37.03 21.17 44.66 803.40
expr min lq mean median uq max
2 rowMedians 1.0000 1.000 1.000 1.000 1.000 1.00
1 colMedians 0.9545 1.429 1.667 1.158 1.693 11.34
Figure: Benchmarking of colMedians() and rowMedians() on integer+10x10 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

100x100 integer matrix

> X <- data[["100x100"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654570 35.0    1168576  62.5  1168576  62.5
Vcells 12129737 92.6   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654564 35.0    1168576  62.5  1168576  62.5
Vcells 12134780 92.6   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on integer+100x100 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 0.1736 0.2156 0.254 0.2575 0.2741 1.206
apply+median 5.1330 7.7029 8.683 7.9037 8.2165 41.385
expr min lq mean median uq max
colMedians 1.00 1.00 1.00 1.00 1.00 1.0
apply+median 29.57 35.73 34.19 30.69 29.98 34.3
Table: Benchmarking of rowMedians() and apply+median() on integer+100x100 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 0.1959 0.2048 0.260 0.2625 0.2983 0.4415
apply+median 6.1620 7.9982 9.009 8.5262 8.9633 22.8178
expr min lq mean median uq max
rowMedians 1.00 1.00 1.00 1.00 1.00 1.00
apply+median 31.45 39.05 34.65 32.48 30.04 51.68
Figure: Benchmarking of colMedians() and apply+median() on integer+100x100 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on integer+100x100 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 173.6 215.6 254 257.5 274.1 1206.4
rowMedians 195.9 204.8 260 262.5 298.3 441.5
expr min lq mean median uq max
colMedians 1.000 1.00 1.000 1.000 1.000 1.000
rowMedians 1.129 0.95 1.024 1.019 1.089 0.366
Figure: Benchmarking of colMedians() and rowMedians() on integer+100x100 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

1000x10 integer matrix

> X <- data[["1000x10"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654605 35.0    1168576  62.5  1168576  62.5
Vcells 12129760 92.6   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654599 35.0    1168576  62.5  1168576  62.5
Vcells 12134803 92.6   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on integer+1000x10 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 0.1763 0.199 0.2288 0.2244 0.2498 0.4581
apply+median 1.2091 1.402 1.5806 1.4717 1.6062 3.9639
expr min lq mean median uq max
colMedians 1.000 1.000 1.000 1.000 1.000 1.000
apply+median 6.858 7.045 6.907 6.557 6.429 8.653
Table: Benchmarking of rowMedians() and apply+median() on integer+1000x10 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 0.1886 0.2006 0.2319 0.2387 0.2491 0.3792
apply+median 1.2946 1.3928 1.4640 1.4266 1.4761 2.5453
expr min lq mean median uq max
rowMedians 1.000 1.000 1.000 1.000 1.000 1.000
apply+median 6.863 6.944 6.315 5.977 5.927 6.713
Figure: Benchmarking of colMedians() and apply+median() on integer+1000x10 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on integer+1000x10 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 176.3 199.0 228.8 224.4 249.8 458.1
rowMedians 188.6 200.6 231.9 238.7 249.1 379.2
expr min lq mean median uq max
colMedians 1.00 1.000 1.000 1.000 1.0000 1.0000
rowMedians 1.07 1.008 1.013 1.063 0.9969 0.8277
Figure: Benchmarking of colMedians() and rowMedians() on integer+1000x10 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

10x1000 integer matrix

> X <- data[["10x1000"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654643 35.0    1168576  62.5  1168576  62.5
Vcells 12130362 92.6   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654637 35.0    1168576  62.5  1168576  62.5
Vcells 12135405 92.6   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on integer+10x1000 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 0.2102 0.2187 0.2578 0.2614 0.2843 0.3661
apply+median 46.2210 58.3153 68.2105 66.3304 73.2646 122.1813
expr min lq mean median uq max
colMedians 1.0 1.0 1.0 1.0 1.0 1.0
apply+median 219.9 266.7 264.6 253.8 257.7 333.7
Table: Benchmarking of rowMedians() and apply+median() on integer+10x1000 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 0.2102 0.2152 0.2588 0.2645 0.2733 0.3588
apply+median 45.9620 53.3136 61.0696 59.6471 63.9272 163.2089
expr min lq mean median uq max
rowMedians 1.0 1.0 1 1.0 1.0 1.0
apply+median 218.7 247.8 236 225.5 233.9 454.9
Figure: Benchmarking of colMedians() and apply+median() on integer+10x1000 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on integer+10x1000 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 210.2 218.7 257.8 261.4 284.3 366.1
rowMedians 210.2 215.2 258.8 264.5 273.3 358.8
expr min lq mean median uq max
colMedians 1 1.0000 1.000 1.000 1.0000 1.00
rowMedians 1 0.9842 1.004 1.012 0.9614 0.98
Figure: Benchmarking of colMedians() and rowMedians() on integer+10x1000 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

100x1000 integer matrix

> X <- data[["100x1000"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654678 35.0    1168576  62.5  1168576  62.5
Vcells 12130708 92.6   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654672   35    1168576  62.5  1168576  62.5
Vcells 12180751   93   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on integer+100x1000 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 1.637 2.025 2.166 2.139 2.205 4.451
apply+median 55.133 72.293 82.732 79.919 90.246 122.047
expr min lq mean median uq max
colMedians 1.00 1.00 1.0 1.00 1.00 1.00
apply+median 33.67 35.71 38.2 37.37 40.92 27.42
Table: Benchmarking of rowMedians() and apply+median() on integer+100x1000 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 1.782 1.813 1.974 1.858 2.094 2.687
apply+median 53.285 67.142 74.188 71.343 77.067 141.831
expr min lq mean median uq max
rowMedians 1.00 1.00 1.00 1.0 1.0 1.00
apply+median 29.91 37.04 37.58 38.4 36.8 52.79
Figure: Benchmarking of colMedians() and apply+median() on integer+100x1000 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on integer+100x1000 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
2 rowMedians 1.782 1.813 1.974 1.858 2.094 2.687
1 colMedians 1.637 2.025 2.166 2.139 2.205 4.451
expr min lq mean median uq max
2 rowMedians 1.000 1.000 1.000 1.000 1.000 1.000
1 colMedians 0.919 1.117 1.097 1.151 1.053 1.657
Figure: Benchmarking of colMedians() and rowMedians() on integer+100x1000 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

1000x100 integer matrix

> X <- data[["1000x100"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654715 35.0    1168576  62.5  1168576  62.5
Vcells 12131126 92.6   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654709   35    1168576  62.5  1168576  62.5
Vcells 12181169   93   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on integer+1000x100 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 1.461 1.625 1.804 1.719 1.986 4.044
apply+median 8.488 11.622 12.755 13.143 13.749 20.217
expr min lq mean median uq max
colMedians 1.00 1.000 1.000 1.000 1.000 1
apply+median 5.81 7.153 7.069 7.647 6.922 5
Table: Benchmarking of rowMedians() and apply+median() on integer+1000x100 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 1.558 1.687 2.105 1.774 2.039 24.85
apply+median 8.628 9.586 12.758 10.865 13.259 139.79
expr min lq mean median uq max
rowMedians 1.000 1.000 1.00 1.000 1.000 1.000
apply+median 5.537 5.683 6.06 6.124 6.503 5.626
Figure: Benchmarking of colMedians() and apply+median() on integer+1000x100 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on integer+1000x100 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 1.461 1.625 1.804 1.719 1.986 4.044
rowMedians 1.558 1.687 2.105 1.774 2.039 24.847
expr min lq mean median uq max
colMedians 1.000 1.000 1.000 1.000 1.000 1.000
rowMedians 1.067 1.038 1.167 1.032 1.026 6.145
Figure: Benchmarking of colMedians() and rowMedians() on integer+1000x100 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

Data type "double"

Data

> rmatrix <- function(nrow, ncol, mode = c("logical", "double", "integer", "index"), range = c(-100, 
+     +100), naProb = 0) {
+     mode <- match.arg(mode)
+     n <- nrow * ncol
+     if (mode == "logical") {
+         X <- sample(c(FALSE, TRUE), size = n, replace = TRUE)
+     }     else if (mode == "index") {
+         X <- seq_len(n)
+         mode <- "integer"
+     }     else {
+         X <- runif(n, min = range[1], max = range[2])
+     }
+     storage.mode(X) <- mode
+     if (naProb > 0) 
+         X[sample(n, size = naProb * n)] <- NA
+     dim(X) <- c(nrow, ncol)
+     X
+ }
> rmatrices <- function(scale = 10, seed = 1, ...) {
+     set.seed(seed)
+     data <- list()
+     data[[1]] <- rmatrix(nrow = scale * 1, ncol = scale * 1, ...)
+     data[[2]] <- rmatrix(nrow = scale * 10, ncol = scale * 10, ...)
+     data[[3]] <- rmatrix(nrow = scale * 100, ncol = scale * 1, ...)
+     data[[4]] <- t(data[[3]])
+     data[[5]] <- rmatrix(nrow = scale * 10, ncol = scale * 100, ...)
+     data[[6]] <- t(data[[5]])
+     names(data) <- sapply(data, FUN = function(x) paste(dim(x), collapse = "x"))
+     data
+ }
> data <- rmatrices(mode = mode)

Results

10x10 double matrix

> X <- data[["10x10"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654758 35.0    1168576  62.5  1168576  62.5
Vcells 12247275 93.5   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654743 35.0    1168576  62.5  1168576  62.5
Vcells 12247403 93.5   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on double+10x10 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 0.0108 0.0192 0.0367 0.0239 0.0562 0.122
apply+median 0.5151 0.7622 0.8520 0.8042 0.8981 1.633
expr min lq mean median uq max
colMedians 1.00 1.0 1.00 1.00 1.00 1.00
apply+median 47.78 39.6 23.22 33.69 15.98 13.38
Table: Benchmarking of rowMedians() and apply+median() on double+10x10 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 0.0162 0.0185 0.0319 0.0262 0.0404 0.0743
apply+median 0.7537 0.7699 0.8571 0.7765 0.9068 2.1796
expr min lq mean median uq max
rowMedians 1.00 1.00 1.00 1.00 1.00 1.00
apply+median 46.62 41.66 26.85 29.66 22.43 29.34
Figure: Benchmarking of colMedians() and apply+median() on double+10x10 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on double+10x10 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 10.78 19.25 36.69 23.87 56.20 122.0
rowMedians 16.17 18.48 31.92 26.18 40.42 74.3
expr min lq mean median uq max
colMedians 1.0 1.00 1.00 1.000 1.0000 1.0000
rowMedians 1.5 0.96 0.87 1.097 0.7192 0.6088
Figure: Benchmarking of colMedians() and rowMedians() on double+10x10 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

100x100 double matrix

> X <- data[["100x100"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654786 35.0    1168576  62.5  1168576  62.5
Vcells 12247286 93.5   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654780 35.0    1168576  62.5  1168576  62.5
Vcells 12257329 93.6   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on double+100x100 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 0.4338 0.5857 0.6407 0.6537 0.6937 0.7437
apply+median 5.2743 6.9086 8.1419 8.4682 8.8922 14.8565
expr min lq mean median uq max
colMedians 1.00 1.0 1.00 1.00 1.00 1.00
apply+median 12.16 11.8 12.71 12.96 12.82 19.98
Table: Benchmarking of rowMedians() and apply+median() on double+100x100 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 0.4319 0.6026 0.6746 0.6858 0.7253 0.7899
apply+median 5.3728 8.4070 8.8032 8.5795 8.8595 21.9304
expr min lq mean median uq max
rowMedians 1.00 1.00 1.00 1.00 1.00 1.00
apply+median 12.44 13.95 13.05 12.51 12.22 27.76
Figure: Benchmarking of colMedians() and apply+median() on double+100x100 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on double+100x100 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 433.8 585.7 640.7 653.7 693.7 743.7
rowMedians 431.9 602.6 674.6 685.8 725.3 789.9
expr min lq mean median uq max
colMedians 1.0000 1.000 1.000 1.000 1.000 1.000
rowMedians 0.9956 1.029 1.053 1.049 1.046 1.062
Figure: Benchmarking of colMedians() and rowMedians() on double+100x100 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

1000x10 double matrix

> X <- data[["1000x10"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654821 35.0    1168576  62.5  1168576  62.5
Vcells 12247991 93.5   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654815 35.0    1168576  62.5  1168576  62.5
Vcells 12258034 93.6   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on double+1000x10 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 0.4115 0.5534 0.5683 0.5626 0.5953 0.9782
apply+median 1.1664 1.7569 1.7738 1.8324 1.9048 3.3938
expr min lq mean median uq max
colMedians 1.000 1.000 1.000 1.000 1.0 1.00
apply+median 2.834 3.175 3.121 3.257 3.2 3.47
Table: Benchmarking of rowMedians() and apply+median() on double+1000x10 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 0.5266 0.6061 0.7037 0.6338 0.7047 2.593
apply+median 1.7423 1.9109 2.3972 2.0306 2.1835 11.630
expr min lq mean median uq max
rowMedians 1.000 1.000 1.000 1.000 1.000 1.000
apply+median 3.308 3.153 3.407 3.204 3.099 4.484
Figure: Benchmarking of colMedians() and apply+median() on double+1000x10 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on double+1000x10 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 411.5 553.4 568.3 562.6 595.3 978.2
rowMedians 526.6 606.1 703.7 633.8 704.7 2593.4
expr min lq mean median uq max
colMedians 1.00 1.000 1.000 1.000 1.000 1.000
rowMedians 1.28 1.095 1.238 1.127 1.184 2.651
Figure: Benchmarking of colMedians() and rowMedians() on double+1000x10 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

10x1000 double matrix

> X <- data[["10x1000"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654859 35.0    1168576  62.5  1168576  62.5
Vcells 12248017 93.5   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654853 35.0    1168576  62.5  1168576  62.5
Vcells 12258060 93.6   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on double+10x1000 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 0.4388 0.4502 0.529 0.4929 0.5892 0.7549
apply+median 47.8101 58.0928 64.496 63.2179 70.1561 90.9257
expr min lq mean median uq max
colMedians 1.0 1 1.0 1.0 1.0 1.0
apply+median 108.9 129 121.9 128.2 119.1 120.4
Table: Benchmarking of rowMedians() and apply+median() on double+10x1000 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 0.4388 0.496 0.5513 0.5049 0.6361 0.7618
apply+median 47.3405 54.822 62.8189 59.3231 66.8636 167.3271
expr min lq mean median uq max
rowMedians 1.0 1.0 1.0 1.0 1.0 1.0
apply+median 107.9 110.5 113.9 117.5 105.1 219.6
Figure: Benchmarking of colMedians() and apply+median() on double+10x1000 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on double+10x1000 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 438.8 450.2 529.0 492.9 589.2 754.9
rowMedians 438.8 496.0 551.3 504.9 636.1 761.8
expr min lq mean median uq max
colMedians 1 1.000 1.000 1.000 1.00 1.000
rowMedians 1 1.102 1.042 1.024 1.08 1.009
Figure: Benchmarking of colMedians() and rowMedians() on double+10x1000 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

100x1000 double matrix

> X <- data[["100x1000"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654894 35.0    1168576  62.5  1168576  62.5
Vcells 12248942 93.5   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654888 35.0    1168576  62.5  1168576  62.5
Vcells 12348985 94.3   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on double+100x1000 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 4.236 4.308 5.027 4.568 5.78 6.672
apply+median 55.728 66.317 75.931 74.960 84.40 131.970
expr min lq mean median uq max
colMedians 1.00 1.00 1.0 1.00 1.0 1.00
apply+median 13.15 15.39 15.1 16.41 14.6 19.78
Table: Benchmarking of rowMedians() and apply+median() on double+100x1000 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 4.521 4.58 5.42 5.548 6.163 6.752
apply+median 57.608 66.22 72.62 71.069 77.162 106.607
expr min lq mean median uq max
rowMedians 1.00 1.00 1.0 1.00 1.00 1.00
apply+median 12.74 14.46 13.4 12.81 12.52 15.79
Figure: Benchmarking of colMedians() and apply+median() on double+100x1000 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on double+100x1000 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 4.236 4.308 5.027 4.568 5.780 6.672
rowMedians 4.521 4.580 5.420 5.548 6.163 6.752
expr min lq mean median uq max
colMedians 1.000 1.000 1.000 1.000 1.000 1.000
rowMedians 1.067 1.063 1.078 1.215 1.066 1.012
Figure: Benchmarking of colMedians() and rowMedians() on double+100x1000 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

1000x100 double matrix

> X <- data[["1000x100"]]
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654931 35.0    1168576  62.5  1168576  62.5
Vcells 12248967 93.5   35610798 271.7 68120027 519.8
> colStats <- microbenchmark(colMedians = colMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 2L, 
+     FUN = median, na.rm = FALSE), unit = "ms")
> X <- t(X)
> gc()
           used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells   654925 35.0    1168576  62.5  1168576  62.5
Vcells 12349010 94.3   35610798 271.7 68120027 519.8
> rowStats <- microbenchmark(rowMedians = rowMedians(X, na.rm = FALSE), `apply+median` = apply(X, MARGIN = 1L, 
+     FUN = median, na.rm = FALSE), unit = "ms")

Table: Benchmarking of colMedians() and apply+median() on double+1000x100 data. The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
colMedians 3.995 4.519 6.009 5.765 6.636 18.78
apply+median 11.581 16.535 18.533 18.478 19.977 34.72
expr min lq mean median uq max
colMedians 1.000 1.000 1.000 1.000 1.000 1.000
apply+median 2.899 3.659 3.084 3.205 3.011 1.849
Table: Benchmarking of rowMedians() and apply+median() on double+1000x100 data (transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.
expr min lq mean median uq max
rowMedians 3.686 3.968 4.473 4.032 4.651 7.419
apply+median 10.706 11.652 14.290 12.639 15.434 25.911
expr min lq mean median uq max
rowMedians 1.000 1.000 1.000 1.000 1.000 1.000
apply+median 2.905 2.937 3.195 3.135 3.319 3.492
Figure: Benchmarking of colMedians() and apply+median() on double+1000x100 data as well as rowMedians() and apply+median() on the same data transposed. Outliers are displayed as crosses. Times are in milliseconds.

Table: Benchmarking of colMedians() and rowMedians() on double+1000x100 data (original and transposed). The top panel shows times in milliseconds and the bottom panel shows relative times.

expr min lq mean median uq max
2 rowMedians 3.686 3.968 4.473 4.032 4.651 7.419
1 colMedians 3.995 4.519 6.009 5.765 6.636 18.776
expr min lq mean median uq max
2 rowMedians 1.000 1.000 1.000 1.00 1.000 1.000
1 colMedians 1.084 1.139 1.343 1.43 1.427 2.531
Figure: Benchmarking of colMedians() and rowMedians() on double+1000x100 data (original and transposed). Outliers are displayed as crosses. Times are in milliseconds.

Appendix

Session information

R Under development (unstable) (2015-02-27 r67909)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1

locale:
[1] LC_COLLATE=English_United States.1252 
[2] LC_CTYPE=English_United States.1252   
[3] LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] markdown_0.7.7          microbenchmark_1.4-2    matrixStats_0.14.0-9000
[4] ggplot2_1.0.0           knitr_1.9.3             R.devices_2.13.0       
[7] R.utils_2.0.0           R.oo_1.19.0             R.methodsS3_1.7.0      

loaded via a namespace (and not attached):
 [1] Rcpp_0.11.4         splines_3.2.0       MASS_7.3-39        
 [4] munsell_0.4.2       lattice_0.20-30     colorspace_1.2-4   
 [7] R.cache_0.11.1-9000 multcomp_1.3-9      stringr_0.6.2      
[10] plyr_1.8.1          tools_3.2.0         grid_3.2.0         
[13] gtable_0.1.2        TH.data_1.0-6       survival_2.38-1    
[16] digest_0.6.8        R.rsp_0.20.0        reshape2_1.4.1     
[19] formatR_1.0.3       base64enc_0.1-3     mime_0.2.1         
[22] evaluate_0.5.7      labeling_0.3        sandwich_2.3-2     
[25] scales_0.2.4        mvtnorm_1.0-2       zoo_1.7-12         
[28] Cairo_1.5-6         proto_0.3-10

Total processing time was 1.73 mins.

Reproducibility

To reproduce this report, do:

html <- matrixStats:::benchmark('colMedians')

Copyright Henrik Bengtsson. Last updated on 2015-03-02 17:09:24 (-0800 UTC). Powered by RSP.

<script> var link = document.createElement('link'); link.rel = 'icon'; link.href = "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACAAAAAgCAMAAABEpIrGAAAADFBMVEX9/v0AAP/9/v3//wBEQjoBAAAABHRSTlP//wD//gy7CwAAAGJJREFUOI3N0rESwCAIA9Ag///PXdoiBk0HhmbNO49DMETQCexNCSyFgdlGoO5DYOr9ThLgPosA7osIQP0sHuDOog8UI/ALa988wzdwXJRctf4s+d36YPTJ6aMd8ux3+QO4ABTtB85yDAh9AAAAAElFTkSuQmCC" document.getElementsByTagName('head')[0].appendChild(link); </script>
Add a custom footer

Benchmark reports

Clone this wiki locally