Interesting application. Glad this was useful to you.

–Loren

I’m trying to find local maxima in an image. Converting to probabilities (rank/numel) and running a local product (colfilt(img,[3 3],’sliding’,@prod)) worked really great,

Roy

Use the dimension input argument for SORT to sort the matrix along the rows or down the columns.

A = magic(3)
sortDownCols = sort(A, 1)
sortAlongRows = sort(A, 2)

Loren wrote about using the dimension input argument with the SUM function about two months prior to this post:

http://blogs.mathworks.com/loren/2007/06/29/sum-things-to-consider/

[m,n] = size(X);
newX = reshape(X,1,m*n);

then use the same thing as loren did with A & B vector

then again reshape my sorted vector back to matrix

this way of sorting is equal to sorting by column then sorting the rows

but if u need to sort only the columns or rows of matrix i dunno how to do it

There is an example in the help to show how to undo the sort. You need to save the indices from the initial sort and use those with the new data.

–Loren

I am trying to do the following:

let’s say i have a matrix ‘A’ of 7000 rows by 50 columns, and i sort using the following:

[sortA, ind] = sort(A,'descend');

now I export this matrix as a .txt file and replace all of the values in Excel with different ones. Then import the altered matrix back into matlab as ‘B’. Therefore ‘B’ has the same dimensions as A but the values have been altered. How can i apply the reverse sort (‘SortA’ to ‘A’) to matrix ‘B’?
Thanks in advance.

Here is one solution:

function idx = USearch(p,v);
%
% Un-Ordered Search for idx : v = p(idx).
% Complexity: O(n);
% USE: for i = 1:7,p(i) = USearch(a,b(i)); end;
% Derek O'Connor, 31 Jan 2011.
%
n = length(p);
idx=0;
for k = 1:n
    if p(k) == v
        idx=k;
        return;             % stops after first find.
    end;
end;
return; % USearch

Use this function as follows:

>> for i = 1:7,p(i)=USearch(a,b(i)); end; disp(p)
     5     1     3     6     7     2     4

Unfortunately this method of finding p(1:n) is O(n^2).

Homework Assignment (Due 7 Jan 2011): Find a quicker method.

Good luck, Derek O’Connor.

see the document
Correlation and aggregate loss distributions with an emphasis on the Iman-Conover Method. Mildenhall, S.J. 2005

Fun question, easy answer if you know what function to use:

a = [1 8 3 17 0 4 7]
b = [0 1 3 4 7 8 17]
[~,ind] = ismember(b,a)

% the output
>> a = [1 8 3 17 0 4 7]
a =
     1     8     3    17     0     4     7
>> b = [0 1 3 4 7 8 17]
b =
     0     1     3     4     7     8    17
>> [~,ind] = ismember(b,a)
ind =
     5     1     3     6     7     2     4

–Loren