how to 7004829X1 into 5000X1 blocks?

I want to separate 1400 blocks and they receive 10 data from the previous and next matrices

[TADA]

Could You Publish An Example Of The Input And Desired Output?

[Cem SARIKAYA]

For example, if I have matrix: A = [1,2,3,4,5,6,7,8,9,10,11,12,13];

and I want matrices:

B = [1,2,3,4,5]; C = [4,5,6,7,8]; D=[8,9,10,11,12];

but i have 7000000x1 matrix and i need 1400 pieces matrix

Answer 1

[Image Analyst]

You say "they receive 10 data from the previous and next matrices" so you want an overlap of 10 when the window jumps to the next location.

You can use blockproc() if you have the Image Processing Toolbox to process those blocks with jumps of whatever you want and a window size of whatever you want. For example a window size of 1400 but the window jumps 1390 each time, instead of 1400, so that you have an overlap of 10 in each window. See attached example and adapt as needed.

[Cem SARIKAYA]

yes this is exactly my teacher says but i do not understand that the attached (blockproc_overlapping.m) would you simplyfy it?

[Image Analyst]

OK, admitted, blockproc() is kind of tricky to understand and use. Perhaps you'd like a simple and intuitive for loop like this:

data = rand(7004829, 1);
overlap = 10;
jumpDistance = 5000 - overlap
for k = 1 : jumpDistance :length(data)
	index1 = k;
	index2 = k + 5000 - 1;
	if index2 > length(data)
		index2 = length(data);
	end
	fprintf('Processing from index %d to index %d\n', index1, index2);
	thisBlock = data(index1:index2);
	% Now do something with this block.
end
fprintf('Done!\n');

If that does the trick, then maybe you could at least "Vote" for the answer.

[Cem SARIKAYA]

this is what i need but this code overwrites all blocks in thisBlock would you change it?

[Image Analyst]

Why do you need to save them? Is there some reason you need them after the loop exits? Why can't you just process them right there in the loop?

If for some reason you need the blocks later, after the loop is done, then you can save them to a cell array. You could save them to a 2-D numerical array but since the last block is only 3859 elements long, you can't do that (unless you wanted to pad the rest with some number).

thisBlock{k} = data(index1:index2);

[Cem SARIKAYA]

thank you very much you correct my perspective. I appreciate.

Answer 2

[Mark Sherstan]

Try this (you will have to adjust for your larger vector) but should do the trick:

A = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15];
count = 1;

for ii = 3:3:length(A)-3
    idxLow = ii-2;
    idxHigh = ii+2;
    out{count} = A(idxLow:idxHigh)';
    
    count = count + 1;
end

[Cem SARIKAYA]

thank you for your answer but this code creates tihis: Columns 1 through 3

{0×0 double} {0×0 double} {1×5 double}

Columns 4 through 6

{0×0 double} {0×0 double} {1×5 double}

Columns 7 through 9

{0×0 double} {0×0 double} {1×5 double}

Columns 10 through 12

{0×0 double} {0×0 double} {1×5 double}

[Cem SARIKAYA]

i want;

{5000x1 double}

{5000x1 double}

{5000x1 double}

{5000x1 double}

{5000x1 double}

{5000x1 double}

.

.

etc

[Mark Sherstan]

Sorry I missed a counter. Correction was made in the original post and located below for your convenience.

A = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15];
count = 1;

for ii = 3:3:length(A)-3
    idxLow = ii-2;
    idxHigh = ii+2;
    out{count} = A(idxLow:idxHigh)';
    
    count = count + 1;
end

out =

  1×4 cell array

    {5×1 double}    {5×1 double}    {5×1 double}    {5×1 double}

[Cem SARIKAYA]

thank you very much but, how can I change it to 5000 values

Answer 3

[TADA]

You can either generate a matrix with each column representing one 1400 part vector as you specified, or generate a cell array, with each cell containing that vector.

a = (1:15)';

% generate a matrix of 5x3
b = reshape(a, [5,3])

b =

     1     6    11
     2     7    12
     3     8    13
     4     9    14
     5    10    15

% generate a cell array of 3 cells with 5x1 vector in each cell
c = mat2cell(a, repmat(5, 1, 3), 1);
celldisp(c);

c{1} =
 
     1
     2
     3
     4
     5

 
 
c{2} =
 
     6
     7
     8
     9
    10

 
 
c{3} =
 
    11
    12
    13
    14
    15
    

as for the large vector you have:

a = (1:7000000)';

b = reshape(a, [1400, 5000]);
c = mat2cell(a, repmat(1400, 1, 5000), 1);

[Image Analyst]

He says he needs (for some reason) "10 data from the previous and next matrices" so is the overlap of 10 elements on the ends of your matrices taken into account with this code? It's not clear since I don't see "10" anywhere in the code. Where in your code do you specify the overlap?

[TADA]

My bad, I missread the whole example...

back to the drawing board...

[Cem SARIKAYA]

thank you very much but i need like this For example, if I have matrix: A = [1,2,3,4,5,6,7,8,9,10,11,12,13];

and I want matrices:

B = [1,2,3,4,5]; C = [4,5,6,7,8]; D=[8,9,10,11,12];

[Image Analyst]

That's an example of overlap of 1. You can use blockproc() if you have the Image Processing Toolbox to process those blocks with jumps of whatever you want and a window size of whatever you want. For example a window size of 1400 but the window jumps 1390 each time, instead of 1400 so that you have an overlap of 10 in each window. See attached example in my Answer.

[TADA]

Thanks Image Analyst for pointing out my mistake.

Here's the rectified version:

% input
a = (1:13);

% settings
overlap = 2;
n = 5;

% generate the indices of column starts
i = 1:(n-overlap):(length(a)-n);

% generate the output matrix
x = bsxfun(@plus, i, (0:(n-1))');

x =

     1     4     7
     2     5     8
     3     6     9
     4     7    10
     5     8    11