how can I create cumulative matrix.

Hello a question please.

I have matrix P with size (452*500). I want to select random selection with roulette wheel selection. I write something:

   r=rand;
   c=reshape(p,1,[]);
   c1=cumsum(c);
   j=find(r<=c1,1,'first');
   [j1 j2]=ind2sub(size(p),j);

end

but I want to remove part " c=reshape(p,1,[]);" and create cumulative matrix of p directly without reshaping it to matrix c. How can I do it?

Thank you.

Answer 1

[Stephen23]

Something like this? :

r = rand;
jj = find(r<=cumsum(p(:)),1,'first');
[j1,j2] = ind2sub(size(p),jj);

Note I also renamed your variable j, which is also one of names of the inbuilt imaginary unit .

[fatema saba]

Thank you,but is there any way to create cumulative matrix of p directly. I mean, for example p is matrix "(500,500). I want to create cumulative mtrix of p for example named pcum that it's size is "(500,500)".

[Stephen23]

This depend on you and what you want your code to do. Read the documentation for cumsum, you will see that it always calculates the sum along one dimension (row, column, etc). Lets have a look at what the documentation says:

• If A is a vector, then cumsum(A) returns a vector containing the cumulative sum of the elements of A.
• If A is a matrix, then cumsum(A) returns a matrix containing the cumulative sums for each column of A.

...

• B = cumsum(A,dim) returns the cumulative sum of the elements along dimension dim. For example, if A is a matrix, then cumsum(A,2) returns the cumulative sum of each row.

So the documentation tells us that if you want a column vector of cumulative sums of each row (or a row of the columns, etc), then the answer to your question is yes, you can do it without reshaping the matrix p first. In every case it is summing only along one dimension, so if you want the cumulative sum over all elements of a matrix, you will have to convert it first to a vector, and probably the neatest way of doing this is using p(:).

[fatema saba]

Thank you for your kindness and clear explanation.