Hello been trying to extract array from a big matrix so because of that taken magic(5) but issue is that when using for loop it works but trying to use in vectorization it gives error

a=magic(5);
c=zeros(3,3,3);
for i=1:3
c(:,:,i)=a(i:3+(i-1),1:3)
end

above code works.

but issue is when using

i=1:3;
c(:,:,i)=a(i:3+(i-1),1:3)

it gives error

Assignment has fewer non-singleton rhs dimensions than non-singleton 
subscripts

a(i:3+(i-1),1:3) is use to extract 3 x 3 matrix

Answer 1

[Walter Roberson]

When you do

i=1:3;
a(i:3+(i-1),1:3)

then you are attempting to use a vector in the base position and a vector in the final position for the colon operator. Look again at https://www.mathworks.com/help/matlab/ref/colon.html#bviscw3-1 and see that those are required to be scalars.

MATLAB does not provide any direct way to do the kind of ragged indexing you want to do.

The approach you need to take in MATLAB is to use sub2ind() or equivalent to construct the linear indexes of the elements you wish to extract, and use linear indexing of the source array.

[ARSALAN RIAZ]

thanks but want to extract multiple elements not a single array!!!

[Walter Roberson]

Example:

a = magic(5);
idx = sub2ind(size(a), [1 1 1; 2 2 2], [1 2 3; 2 3 4]);
a(idx)

ans =
    17    24     1
     5     7    14
>> a
a =
    17    24     1     8    15
    23     5     7    14    16
     4     6    13    20    22
    10    12    19    21     3
    11    18    25     2     9

If your index array were 3D where I show the 2D indices, then you would get a 3D output.

[ARSALAN RIAZ]

thanks that make some sense!!! if i want to extract multiple submatrix such as if orignal matrix is 5x5 and i want sub-matrix from 1 row to 3 row,2 row to 4 row,3 row to 5 row. colums are same i.e 1 to 3 in every sub-matrix.sub-matrix are of size 3x3 and save the sub matrix in array of cell. take take sub-matrix one by one and perform operation on every submatrix

[Walter Roberson]

See also the code at https://www.mathworks.com/matlabcentral/answers/348500-extracting-last-non-nan-200-columns-from-matrix-with-varying-number-of-nans-ending-the-rows#answer_273971

[ARSALAN RIAZ]

well thanks. is therte any way to Multiplying each matrix column by a column vector without using bsxfun? purpose is to make code faster

[Walter Roberson]

https://www.mathworks.com/matlabcentral/fileexchange/25977-mtimesx-fast-matrix-multiply-with-multi-dimensional-support