How to Obtain a Portion of an n-dimensioned Matrix

I have a 6 dimensional matrix with dimensions 16x20x22x6x3x13. I identified specific elements of that matrix with the following:

y = find(spec_neg(:,:,:,:,:,1) ~= 0);

Since y is now the indices of the spec_neg and I would like these in matrix subscript form, I determined the subscripts by doing the following:

[i,j,k,l,m,n] = ind2sub([16,20,22,6,3,13],y);

and then concatenated these subscripts into matrix form:

y_ind = cat(2,i,j);
y_ind = cat(2,y_ind,k);
y_ind = cat(2,y_ind,l);
y_ind = cat(2,y_ind,m);
y_ind = cat(2,y_ind,n);

I want to obtain the values of spec_neg that correspond to the subscript values of y_ind but when I do the following:

spec_neg(y_ind)

I get all zeros so that is obviously not the values in y.

What am I doing wrong?

[Matt J]

Since y is now the indices of the spec_neg and I would like these in matrix subscript form

Be careful. The vector y does not contain indices of spec_neg. It contains indices of the 5-dimensional array spec_neg(:,:,:,:,:,1). The two would not coincide if you were indexing a different section like spec_neg(:,:,:,:,:,10).

So, which is it that you want?

Answer 1

[Matt J]

I would extract the values prior to extracting the indices, as below. Notice also that you don't have to do all those painful multiple calls to cat(2,...):

      s=spec_neg(:,:,:,:,:,1);
      idx=logical(s);

      values=s(idx);

      [c{1:5}]=ind2sub(size(s), find(idx));

      y_ind=cat(2,c{:});

Notice also that, while spec_neg is 6-dimensional, spec_neg(:,:,:,:,:,1) is only 5-dimensional, so I think you really want 5 columns in y_ind.

[Jim]

I had to take a little time to think on this. I see now that spec_neg(:,:,:,:,:,1) is only 5 dimensional. I didn't think about it when I did this originally.

[Jim]

I'm having a hard time understanding how parts of the last two lines work in your solution. I understand the ind2sub and cat functions but I don't understand how the bracket and brace operators work. Can you help me understand them?

[c{1:5}]=ind2sub(size(s), find(idx));
y_ind=cat(2,c{:});

[Sean de Wolski]

Brackets and braces allow for "comma-separated list expansion" both on inputs and outputs:

http://www.mathworks.com/help/releases/R2014b/matlab/matlab_prog/comma-separated-lists.html

There's a learning curve but very powerful and arguable necessary for working with n-dimensional arrays.

[Jim]

In the above example, I want to obtain the 13 values in the 6th dimension of spec_neg that correspond to the indices in y_ind. Why does the following not work?

spec_neg(y_ind)

I would also like those values concatenated with y_ind to yield a 2 dimensional matrix.

[Matt J]

Why does the following not work?

Because y_ind is a concatenation of subscripts, which is not a legal kind of indexing in MATLAB. Only when y_ind is a linear or logical index array can it be used to pull out data from an array with the single argument syntax,

    spec_neg(y_ind)

Subscripts must be used in a comma-separated syntax

    spec_neg(I,J,K,...)

Furthermore, only rectangular sub-matrices and sub-arrays can be extracted with subscripts. To extract an arbitrary list of values, you must again use logical or linear indexing.