Special matrix without for loop (vectorizing loop)

Hi Guys

I am trying to improve a performance of a code which develops a special matrix. This matrix looks really simple but I couldn't find a solution to develop it without a for loop. This is a huge matrix and a for loop causes the code takes hours to run.

imagine a vector A = [ 1; 2; 3; 4; 5; 6; 7 ] I want to develop matrix B based on A. numbers are indexes not real numbers.

B = [1 2 3 4; 2 3 4 5; 3 4 5 6; 4 5 6 7 ]

your help is really appreciated.

Regards Mitch

Answer 1

[Andrei Bobrov]

variant 1

 B = hankel(A(1:4),A(4:end))

variant 2

B1 = cumsum([A(1:4) ones(4,3)],2)

variant 3

B2 = bsxfun(@plus,A(1:4),[0 1:3])

[M]

Wow

Saved me hips of time

Thanks :)

[Oleg Komarov]

If you think that this answer soved your problem, please accept it.

[M]

about variant 2 3 they are good when the numbers are integers like 1 2 3 4 5 6 ... but the numbers are just indexes of my matrix not real numbers. So I think only Hankel is working properly. I developed my code again based on Hankel and I noticed the Hankle even deteriorates the speed of code in comparison to normal for-loop code. I am talking doing the same Hankle function on a vector with 200,000 elements for 1000 times.

hankel

Hankel matrix

Syntax

H = hankel(c)

H = hankel(c,r)

in which my c has 150 elements.

[Oleg Komarov]

Fo variants 2 and 3 then use A(B1), A(B2);