I set up 2 problems as follows:
I have two matrices (Mat1 and Mat2). Both matrices are of equal size. I have four output matrices (Output1, Output2, Output3, Output4 respectively) both the same size as Mat1 and Mat2.
Problem 1:
In Mat2, Identify the row that contains the maximum value in column1. Lets assume this is row 1.
Go to Row 1 of Mat1 and extract the first 3 columns of Mat1 Row1 and store in Output1. Store all other rows in Mat2 for the first 3 columns. At this stage Output1 is 1x3. Output 2 is (n-1)x3.
Move to Column 4 of Mat2. Identify the row that contains the maximum value. Lets say this is row 5.
Go to row 5 of Mat1 column 4. Store Row5 columns 4,5,6 in Output1. Store all other rows of Mat1 for columns 4,5,6 in Output2. ... Repeat this process for all columns in Mat2 following the sequence:1,4,7,9 etc. In this case, i have 25 columns for Mat1 and Mat2, so the sequence will end at 24.
I need to be able to change the sequence from 1,4,7,9 etc, to 1,13,25 etc.
Problem2: is equivalent to problem 1, except this time i identify the rows that contain the top-two values in every stage.
In Mat2, Identify the rows that contain the top-two values in Column1. Lets say these rows are 2 and 5. Store the first 3 columns of rows 2 and 5 of Mat1 in Output3. Store all remaining rows (column 1-3) of Mat1 in Output4.
Move to Column 4 of Mat2. Identify the rows which contain the top two values in column 4 of Mat2. lets say rows 1 and 2.
Move to Column 4 of Mat 1. Store column 4,5,6 Row 1 and 2 into Output3. Store all remaining rows in Output4.
Sidenote: This process must be easily extended for matrices with 1000x1000 dimensions. So would prefer not to do this manually.
Mat1 <- data.frame(matrix(nrow = 10, ncol =25, data = rnorm(250,0,1)))
Mat2 <- data.frame(matrix(nrow = 10, ncol =25, data = rnorm(250,0,1)))
> Mat1
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13
1 -2.22415466 0.98712728 1.0084356 0.58447183 0.2608830 -0.341029099 -0.71693894 -0.61653058 -0.24790470 0.10777970 -1.68562271 -1.6638535 -0.5538468
2 1.11444365 -0.34865955 0.7518822 -0.07573724 0.1336811 -0.831275643 -0.15564822 -0.68849375 -0.05094047 0.21990082 -0.69879135 -0.6348292 1.0172304
3 -0.05367747 0.08654206 -0.3023270 -0.67335942 -1.1173279 0.004670625 0.52482501 0.78330982 1.18795853 -0.06513613 0.42353439 -0.4152209 1.7174158
4 0.42118984 -0.43257583 -1.3368036 1.64849798 0.8294276 1.256987496 -0.50440892 1.07686292 0.94196135 2.90916270 -0.08714083 0.1094395 1.1715895
5 -0.13720451 -0.94864452 1.9751962 -0.70523555 0.1431405 0.569928767 0.54877505 -0.44571903 -1.16282161 -1.65590032 -0.17710859 -0.8904316 0.3252576
6 0.64336424 -0.38277541 -1.6512377 -0.06542054 -0.1195322 0.666255832 0.60826054 1.88822842 -0.52952627 -0.44776682 0.04321836 -0.6190585 -0.9529690
7 -1.04160098 1.10952094 -0.9186759 0.77437293 -0.2284926 -0.113106151 -0.32092624 1.34157301 2.33813068 1.21812714 0.13165646 0.5532299 -1.3470645
8 1.22940987 -1.26271164 -1.2483658 -2.00578793 -0.6773794 -0.228135998 -0.06223206 -1.97606848 1.67339247 -0.47268196 -0.83544561 -0.3313278 -0.2373613
9 0.08485706 -1.60594589 0.8549923 -0.23394708 -0.5978692 -0.321839877 -0.55298452 -0.08387815 -0.99196489 0.83364114 -0.19579612 -0.8017648 -0.2238073
10 -1.71702699 0.39086484 -0.9974210 0.86232862 -0.2755329 -0.160656438 0.49669949 0.73763073 -0.42380390 1.91208332 -0.27778479 0.7866471 0.1813511
> Mat2
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13
1 0.11053732 -0.5750170 2.58105259 -1.6895285 0.0508918 -0.54188929 -0.92292169 1.4972970 0.009239807 -0.1706461 -0.8942262 -1.6351505 0.2029262
2 -0.83802776 -0.9322157 -0.34753884 0.8164819 0.7318198 0.09162218 0.15971493 -2.6731067 1.554323641 -0.3161967 0.4622101 -1.9521229 -1.3229961
3 0.61368153 -1.3650360 0.95674229 0.4582117 -0.6959545 -0.59627428 1.94172156 1.6784237 -0.482524695 -0.0514944 -0.4608930 -0.5456863 -0.1340540
4 -1.03156503 -0.2516495 0.76770177 -0.7841354 -3.2404904 -1.76276859 1.57421914 0.9782458 -1.364451438 -0.6437429 0.7485424 -0.8778284 1.7587504
5 0.01183232 0.6825633 1.39634308 1.4136879 0.5166420 0.76930390 0.67210932 1.3007904 -0.284451411 0.5163457 0.3198626 0.8030497 -1.4320064
6 -0.06110883 -0.6762991 0.56105196 0.9767543 -1.0016294 -0.84811626 -0.83319744 -1.1777865 -1.185631394 -0.5673733 0.2956725 0.5425602 -1.0510479
7 -0.56195630 1.3883881 0.09995573 0.6722959 -1.6205290 0.32085867 -0.94243554 -0.2340429 -1.299085265 -0.4433517 0.4424583 -2.8887970 0.1679859
8 1.04612102 0.8360530 0.07005306 0.4818317 1.1857504 0.13649605 1.35261983 0.8008935 -0.101922164 0.6773003 -1.0265770 0.1859912 0.2678461
9 0.88419676 -1.7012899 -1.09656000 -0.4360276 0.6238451 -2.03256276 -1.12575579 1.8407234 0.522372401 -0.6229582 0.6727720 -0.5695190 0.6298388
10 -0.68648649 -0.6689894 -0.56849261 -1.9012760 1.1418180 0.46377789 -0.08107475 1.4378120 -1.489367198 -0.7682887 -0.2858680 0.9584056 1.3178700
So for example:
which.max(Mat2[,1]) # 8
so go to row 8 of Mat1 and store the first 3 cols in Output1.
Output1[1, 1:3] # 1.22940987 -1.26271164 -1.2483658
Store all other rows of Mat1 for cols 1 to 3 in Output2.
which.max(Mat2[, 4]) # 5
implies
Output1[1, 4:6] # -0.70523555 0.1431405 0.569928767
And so on and so forth.
Does this do what you need?
Juggle <- function(m1, m2, col, step, top_n = 1) {
if (length(col) == 1) {
output_rows <- sort(m2[, col], index.return = TRUE, decreasing = TRUE)$ix
col_idx <- seq(col, col + step - 1)
top_idx <- seq(top_n)
bottom_idx <- seq(top_n + 1, nrow(m2))
max_out <- m1[output_rows[top_idx], col_idx]
rest_out <- m1[output_rows[bottom_idx], col_idx]
return(list(max_out = max_out, rest_out = rest_out))
} else {
left <- Juggle(m1, m2, col[1], step, top_n)
right <- Juggle(m1, m2, col[seq(2, length(col))], step, top_n)
return(list(
max_out = cbind(left$max_out, right$max_out),
rest_out = cbind(left$rest_out, right$max_out)
))
}
}
m1 and m2 are the matrices or data frames.
col is a single or sequence of columns like 1, 4, 7, 9 or 1,13,25.
step is the amount of columns to extract.
top_n is the amount of rows to extract.
This will return a list with components max_out for the top_n row extracts and rest_out for the N-top_n row extracts.
Collected from the Internet
Please contact [email protected] to delete if infringement.
Comments