Max()-function and multiple dimensions

Hello,

I'm trying to maximize a matrix that is of several dimensions. I also need to know the locations of the maximized value, which is actually the problem. How do I go about doing this in the most simple way?

So, for example, x=rand(10,10,10,10,10); Of course, by simple max(max(max(max(x) I could get the maximal value of x. How about the location? (for example the maximum could be located in x(1,2,3,2,4))

Cheers,

M

[Image Analyst]

I think you're getting only answers on how to locate the max value because no one knows what you want to do to maximize the array. What is the array dependent on such that changing those parameters will maximize the array? And what does it mean to maximize an array, as opposed to a single value. Do you want to vary some parameter, say lambda or whatever you want to call it, such that the sum, or mean, of all the values in the array is maximized?

Answer 1

[Ahmet Cecen]

Do:

 [C I]=max(x(:));

now C is the maximum value, I is the index of where it is, so that x(I) is C.

If you want the 5D coordinates explicitly, then read ind2sub.

[Jan]

E.g.:

[cPos{1:ndims(x)}] = ind2sub(size(x), I);
Pos = [cPos{:}];

[Shivanand Venkanna Sheshappanavar]

what if the matrix is a 2D matrix ? We need maxval, rowindex and columnindex

[Steven Lord]

The solution Ahmet and Jan posted works fine for 2-D matrices, 5-D arrays, etc.

Answer 2

[Azzi Abdelmalek]

Edit

A=rand(10,10,10,10,10);
[max_value,idx]=max(A(:));
q=['[' sprintf('x%d,',1:ndims(A))];
q(end)=']';
eval([q '=ind2sub(size(A),idx)'])
eval(['x=' q])

[Jan]

This is cruel. There is no reason to go the indirection over the evil eval. eval is prone to bugs, hard to debug, disables the JIT acceleration and is a severe drawback for efficient programming.