how to find the next maximum value in column of the matrix which is different at the index of the first

Hi!

I have an N * M matrix, i want to find the maximum value in each column,if the index of the first max value is the same with the others,i will find the next maximum value in those columns.

SINR =

0.3170 0.1029 0.0780 0.0014 0.0303

0.0666 0.1213 0.2206 0.4729 0.0542

0.0157 0.0109 0.2463 0.2058 0.5439

0.0404 0.1542 0.1665 0.0251 0.1449

0.0448 0.0264 0.0024 0.0532 0.0341

0.1038 0.0240 0.0028 0.1512 0.0965

0.1099 0.0503 0.2834 0.0706 0.0887

0.0324 0.2983 0.0452 0.0719 0.0344

0.4190 0.4786 0.1560 0.0094 0.0996

0.0868 0.0037 0.0024 0.1878 0.1287

SINRmax =

0.4190 0.4786 0.2834 0.4729 0.5439

0.3170 0.2983 0.2463 0.2058 0.1449

0.1099 0.1542 0.2206 0.1878 0.1287

idx1 =

9 9 7 2 3

1 8 3 3 4

7 4 2 10 10

as we see the index of the max value in column 1 is the same in column 2 ,in this case i will choose the second max value (index 8) for column 2 .

the output : put only the max value of each column and all other values at zero

0 0 0 0 0

0 0 0 0.4729 0

0 0 0 0 0.5439

0 0 0 0 0

0 0 0 0 0

0 0 0 0 0

0 0 0.2834 0 0

0 0.2983 0 0 0

0.4190 0 0 0 0

0 0 0 0 0

there is a function that allows me to do that?

can you help me please

Answer 1

[Dana]

I'm not aware of any built-in function. You'll probably have to write one yourself. Something like:

function [mxs,mxinds] = mymax(A)

    N = size(A,1);
    M = size(A,2);
    
    % allocate arrays for output
    mxs = zeros(M,1);
    mxinds = zeros(M,1);
    
    % This vector will track the indices of rows that we want to consider.
    % Each time we find a max in a row as we go through the columns, we'll
    % remove that index from 'inds'.
    inds = (1:N).';
    
    for j = 1:M     % for each column
        % Find the maximum value in column j among only the rows in inds.
        % Load that maximum value into mxs(j), and the corresponding
        % location into the variable indj.
        [mxs(j),indj] = max(A(inds,j)); 
        
        % Note that indj is the location of the max within the vector
        % A(inds,j), NOT the location within A(:,j). The location within
        % A(:,j) is given by the following:
        mxinds(j) = inds(indj);
        
        inds(indj) = []; % delete that row index from inds
    end
end

This function sequentially looks for maximums of each column starting from column 1, but where for column j we only consider rows that we have not previously identified as a maximum in any of columns 1 to j-1. I'm not sure if this is exactly what you're after. For example, if you only want to do this for columns 1 and 2, you'd need to make a modification. Or if you only wanted to exclude the row a maximum appeared in in the previous column (rather than all previous columns), you'd again have to modify things.

[Ideth Kara]

thanks for your reply,but i want to do this for all the columns and the output should be a matrix representing only the max value

[Dana]

I don't understand what you mean. The code I posted does find the max for every column (subject to the constraint of not considering rows that have already been a maximum). The max values are collected in the vector mxs, so that mxs(j) is the max for column j.

[Dana]

I guess maybe you're confused since mxs and mxinds are column vectors? If you'd prefer row vectors, just change the allocation statements to:

    mxs = zeros(1,M);
    mxinds = zeros(1,M);

[Ideth Kara]

yes i know,but at the output I want to see a matrix contains the max values in different rows not a vector

[Dana]

I still don't get how the function I posted is giving you the wrong answer. So after making the above change to convert mxs and mxinds to row vectors:

SINR = [0.3170    0.1029    0.0780    0.0014    0.0303
    0.0666    0.1213    0.2206    0.4729    0.0542
    0.0157    0.0109    0.2463    0.2058    0.5439
    0.0404    0.1542    0.1665    0.0251    0.1449
    0.0448    0.0264    0.0024    0.0532    0.0341
    0.1038    0.0240    0.0028    0.1512    0.0965
    0.1099    0.0503    0.2834    0.0706    0.0887
    0.0324    0.2983    0.0452    0.0719    0.0344
    0.4190    0.4786    0.1560    0.0094    0.0996
    0.0868    0.0037    0.0024    0.1878    0.1287];

[mxs,mxinds] = mymax(SINR)

mxs =
        0.419       0.2983       0.2834       0.4729       0.5439
mxinds =
     9     8     7     2     3
     

So mxinds is giving you a 9 for column 1 and an 8 for column 2, which I thought is what you're expecting. Which output here exactly is wrong?

[Ideth Kara]

this function still give the same index ,another example after using this function

mxs =

0.8466 0.3992 0.4339 0.4067 0.3415

mxinds =

1 10 4 6 10

the index 10 shouldn't be repeated

[Dana]

I can't reproduce this behavior. I just randomly generated 100,000 matrices of the same dimensions as your SINR, applied the mymax function to each, and not once did I get a repeated index. What input matrix did you use to get the result you showed?

[Ideth Kara]

SINR =

0.8466 0.2115 0.0385 0.0243 0.2741

0.0216 0.3572 0.2595 0.1917 0.2854

0.0375 0.0141 0.0104 0.0860 0.0761

0.0354 0.0680 0.4339 0.0112 0.0443

0.0274 0.0194 0.0419 0.0347 0.0201

0.2997 0.0372 0.0238 0.4067 0.0588

0.0204 0.1418 0.0560 0.0815 0.0075

0.1020 0.0018 0.0871 0.0704 0.0091

0.0085 0.0189 0.2152 0.0321 0.1154

0.0773 0.3992 0.0755 0.2990 0.3415

mxs =

0.8466 0.3992 0.4339 0.4067 0.3415

mxinds =

1 10 4 6 10

i want to save the max value in column 2 (index 10) ,but for the column 5 ,i have to find the next max value in this column which is in (index 9) in order to remove the repetition

the matrix that i want to have shows only the maximum value in each column with different indexes

[Dana]

When I use that value of SINR, I get

mxs =
       0.8466       0.3992       0.4339       0.4067       0.2854
mxinds =
     1    10     4     6     2     

Are you sure you copied and pasted the function correctly?

Also, for column 5, I think index 2 is the correct answer, not 9 (0.2854 > 0.1154).

[Ideth Kara]

it works well ,thank you so much