How to modify the following code

good , using the following code I want to create all the substrings of a sequence s . Eg

s = [ 1 1 1 0 0 1 0 1 ]

using the following code , where n = length substrings :

if true

    % code

N = length ( s ) ; % number rows

n = 4 ;

 A = Hankel(1: N -n +1 , N- n +1: N);

k = 0: n-1;

idx = [ ] ;

for ii = 1 : size ( A, 1 )

    p = A ( ii , :) ;

    while p (end , end) + k ( end) < = N

        p = [p , p ( end , :) + k] ;

    end

    idx = [ idx , p ] ;

end

[ j1 , j2 , j2 ] = unique ( s ( idx ) , ' rows ');

end

The code returns me an array with the result of combinations in this sequence:

0 0 1 0

0 1 0 1

1 0 0 1

1 0 1 1

1 1 0 0

1 1 1 0

My question is this : How I can change the code so that instead of a single sequence s me , I make the rows of a matrix. For example , I want to calculate the substrings of length 4,5 ... n of the rows of the following matrix :

1 1 1 0 0 1 0 1

0 1 1 0 0 0 0 0

and that all combinations the same length of both rows store in a single array , using the above code . How do? thank you very much

[Roger Stafford]

It is important that you carefully define what you mean by a "substring" and show us in detail how that accounts for the six rows of your result. Apparently they must in some sense be contiguous elements of s but to account for the 1 0 1 1 string it is necessary to expand s with another copy of itself. But then why aren't 0 1 1 1 and 1 1 1 1 allowed? I hope you won't define a "substring" as whatever result is given by your code.

[FRANCISCO]

okei, I define the substring according idx. The way to create the substring is by the number of combinations as follows (numbers in parentheses indicate the order);

  s = [1 (1) 1 (2) 1 (3) 0 (4) 0 (5) 1 (6) 0 (7) 1 (8)];

then when you run the above code, the array idx shows me the order:

1 2 3 4

1 3 5 7

2 3 4 5

2 4 6 8

3 4 5 6

4 5 6 7

5 6 7 8

that by using the code:

if true

    % code

[j1, j2, j2] = unique (s (idx), 'rows');

end

I estimated subsequences previous sequence:

0 0 1 0

0 1 0 1

1 0 0 1

1 0 1 1

1 1 0 0

1 1 1 0

I do this for a sequence s, but now I have an array with multiple sequences, for example:

1 1 1 0 0 1 0 1

0 1 1 0 0 0 0 0

and save the result of the calculation of both sequences distinct array to another array

Many thanks

[FRANCISCO]

I can not find solution

[Roger Stafford]

I should point out that the line with

 unique(s(idx),'rows');

will not work as it stands. You would need to write

 unique(reshape(s(idx),n,[])','rows')

to get the results you claim.

Also if the length(s) is sufficiently large that the while-loop executes twice in succession, then the line

 p = [p,p(end,:)+k];

will fail. It should be replaced by

p = [p,p(end,end-n+1:end)+k];

Answer 1

[Roger Stafford]

To use more than one row in s, in place of N = length(s) you can do this:

 [M,N] = size(s);

Then use your code (hopefully corrected in the while-loop as I recommended earlier) to create the row vector idx. Then do this:

 m = length(idx);
 idx = repmat(idx,1,M)+reshape(repmat(0:N:(M-1)*N,m,1),1,[]);
 s = s';
 j1 = unique(reshape(s(idx),n,[])','rows');

[FRANCISCO]

Very good,

and realize the code as follows:

use the following matrix:

1 0 0 1 1 1 0 0

0 0 0 0 1 0 0 1

if true

    % code
[M,N] = size(s);;%numero filas

n = 4;

A = hankel(1:N-n+1,N-n+1:N);

k = 0:n-1;

idx = [];

for ii = 1:size(A,1)

    p = A(ii,:);

    while p(end,end) + k(end) <= N

        p = [p,p(end,end-n+1:end)+k];

    end

    idx=[idx;p];

end

m = length(idx);

idx = repmat(idx,1,M)+reshape(repmat(0:N:(M-1)*N,m,1),1,[]);

s = s';

j1 = unique(reshape(s(idx),n,[])','rows');

end

and Matlab gives me error:

Error using vertcat CAT arguments dimensions are not consistent.

I'm doing wrong?

[FRANCISCO]

I've also tried placing it in the following way:

if true

    % code

[M,N] = size(s);;%numero filas

n = 4;

A = hankel(1:N-n+1,N-n+1:N);

k = 0:n-1;

idx = [];

for ii = 1:size(A,1)

    p = A(ii,:);

    while p(end,end) + k(end) <= N

        p = [p,p(end,end-n+1:end)+k];

    end

    m = length(idx);

    idx = repmat(idx,1,M)+reshape(repmat(0:N:(M-1)*N,m,1),1,[]);

end

end

But matlab, the end of the loop gives me error:

"Error using + Matrix dimensions must agree."

[FRANCISCO]

I also I can the following code to reform without p, but neither was able to calculate for an array:

if true

    % code

N = length(s);

n = 4;

A = hankel(1:N-n+1,N-n+1:N);

k = 0:n-1;

c = ceil((N - A(:,end) + 1)/k(end));

i2 = cumsum(c);

i1 = i2 - c + 1;

idx = zeros(i2(end),n);

for jj = 1:N-n+1

    idx(i1(jj):i2(jj),:) = bsxfun(@plus,A(jj,:),(0:c(jj)-1)'*k);

end

[j1,j2,j2] = unique(s(idx),'rows')

end

Many thanks

[Roger Stafford]

You made one change in your original code and that is apparently what is causing you the trouble. In your original code you had:

 idx = [idx,p];

just after you exit the while-loop and just before the end of the for-loop. However, in the more recent version which failed with the error message "Error using vertcat CAT arguments dimensions are not consistent", you had written instead:

 idx = [idx;p];

with a semicolon instead of a comma. That would invalidate the coding which I suggested to you. My assumption was that idx would emerge from the for-loop as a vector with only a single row, as you had it originally. I recommend you replace the semicolon with a comma and try it out again.