## Sorting from the maximum value to minimum value in a vector

### mr mo (view profile)

on 29 Nov 2017
Latest activity Answered by Image Analyst

on 30 Nov 2017

### Kaushik Lakshminarasimhan (view profile)

Hi. I have a vector named A, e.g.
A=[2 5 8 7 8 9 3 7 6 5 4 1]
I want to sort the members of vector A from maximum to minimum values.
Also I want to find the indices of equal members in the sorted version of vector A.
How can I do that?

### Kaushik Lakshminarasimhan (view profile)

on 29 Nov 2017

A=[2 5 8 7 8 9 3 7 6 5 4 1];
Asorted = sort(A,'descend'); % sorted from max to min
[~,indx] = unique(Asorted);
equalpairs = [(setdiff(1:length(Asorted),indx)-1)' setdiff(1:length(Asorted),indx)']; % each row contains pair of indices with equal values

mr mo

### mr mo (view profile)

on 29 Nov 2017
Thanks a lot. This is what I want.

### Matt J (view profile)

on 29 Nov 2017

Using the UNIQUE command.

mr mo

### mr mo (view profile)

on 29 Nov 2017
Can you write the code ? Thanks a lot.
mr mo

### mr mo (view profile)

on 29 Nov 2017
In the unique command I lost some of the members and I don't want that.
mr mo

### mr mo (view profile)

on 29 Nov 2017
I want to reach this:
sortedA = [9 8 8 7 7 6 5 5 4 3 2 1]
and I want the indices of equal members in the sorted version of A, e.g.
sortedA(2)=sortedA(3)
sortedA(4)=sortedA(5)
sortedA(7)=sortedA(8)

### Image Analyst (view profile)

on 30 Nov 2017

With the accepted solution, if A is
A =[22 25 28 27 28 29 23 27 26 25 22 22 24 21 22]
The accepted solution gives
equalpairs =
2 3
4 5
7 8
11 12
12 13
13 14
However if you use my solution:
A = [2 5 8 7 8 9 3 7 6 5 2 2 4 1 2] + 20
sortedA = sort(A, 'descend')
props = regionprops(sortedA, sortedA, 'MeanIntensity', 'PixelIdxList');
missingNumbers = isnan([props.MeanIntensity]);
props = props(~missingNumbers); % Get rid of nans that occur when there is a missing number.
% Done! Now print them out so we can see:
for k = 1 : length(props)
fprintf('The indexes for %d are ', props(k).MeanIntensity);
fprintf('%d, ', props(k).PixelIdxList);
fprintf('\n');
equalIndexes{k} = [props(k).PixelIdxList];
end
celldisp(equalIndexes)
You will get a cell array where each cell has the indexes for a certain number:
A =
22 25 28 27 28 29 23 27 26 25 22 22 24 21 22
sortedA =
29 28 28 27 27 26 25 25 24 23 22 22 22 22 21
The indexes for 21 are 15,
The indexes for 22 are 11, 12, 13, 14,
The indexes for 23 are 10,
The indexes for 24 are 9,
The indexes for 25 are 7, 8,
The indexes for 26 are 6,
The indexes for 27 are 4, 5,
The indexes for 28 are 2, 3,
The indexes for 29 are 1,
equalIndexes{1} =
15
equalIndexes{2} =
11
12
13
14
equalIndexes{3} =
10
equalIndexes{4} =
9
equalIndexes{5} =
7
8
equalIndexes{6} =
6
equalIndexes{7} =
4
5
equalIndexes{8} =
2
3
equalIndexes{9} =
1
I think having a complete list is probably more useful and convenient than having a list of pairs which doesn't even include all pairs. For example the accepted solution does not list (11, 14), (12,14), etc. as pairs.