matrix from arrays of different size

Hi,

I have a for loop in which i solve a system of ode's. The solution looks fine when I plot however I need to save the data at each iteration in one single matrix.

The code is below:

for q=1:20

        tspan1=q:21;
        [s x] = ode23s(@odebatch,tspan1, [q n0(q) 0],[], h,gr) ;       
         f=x(:,3);

% xmat(:,q)=padcat(x(:,3)) ;

% u=x(:,1); % % hold on % figure(1) % % plot(f,u)

       end

As you can see the time over which the solution to the system of ode is to be found is changing. That means i have arrays of different sizes at each iteration. I want to have something like:

0 NAN 1 0 2 1 3 2 4 3

Any ideas how to do this?

Answer 1

[Cedric]

I don't understand where

 0 NAN 1 0 2 1 3 2 4 3

is coming from, but why do you need to have results in a single array? If you just want to store them, use a cell array, e.g.

 results = cell(20, 1) ;
 for q = 1 : 20
    tspan1 = q : 21 ;
    [s, x] = ode23s(@odebatch, tspan1, [q, n0(q), 0], [], h, gr) ;
    results{q} = [s, x] ;
 end

If you really needed to have them in the same array, you could go for something like (not tested):

 xmat = nan(21, 20) ;    % or xmat = zeros(21, 20) ;
 for q = 1 : 20
    tspan1 = q : 21 ;
    [s, x] = ode23s(@odebatch, tspan1, [q, n0(q), 0], [], h, gr) ;
    xmat(q:21,q) = x(:,3) ;
 end

[Graham]

Hi, Sorry that 0 NAN 1 0 2 1 3 2 4 3 given above was a mistake. It came out like that when I typed it up. Basically I wanted to say that I needed the cells at certain times to have nan. I have altered the second piece of code above. It is nearly doing what I want except it only outputting the first and last solution of the loop in the matrix xmat.

xmat = nan(i, q) ; % or xmat = zeros(i, q) ; for q = 1 : 500:i tspan1 = q : i ; [s, x] = ode23s(@odebatch, tspan1, [q, n0(q), 0], [], h, gr) ; xmat(q:i,q) = x(:,3) ; end

any ideas as to why this would be happening?

[Cedric]

Starting with a side note, be careful with i and j, they are "reserved" (even if everybody uses them) to complex notation. We usually use ii and jj instead for indexing.

The reason your code doesn't work is that you are using a q that goes from 1 to an upper boundary by steps of 500, for indexing columns in xmat (whereas column indices should not evolve by steps of 500). You should have something like the following instead, where the loop index can be used for indexing columns and picking the relevant lower boundary:

 uBnd  = ... ;                           % Upper boundary.
 lBnds = 1 : 500 : uBnd-1 ;              % Lower boundaries (-1 ?).
 n = length(lBnds) ;                     % Number of lBnds and runs.

 xmat = nan(uBnd, n) ;                   % Prealloc/fills xmat with nans.

 for k = 1 : n
    lBnd   = lBnds(k) ;                  % Formerly named 'q'.
    tspan1 = lBnd : uBnd ; 
    [s, x] = ode23s(@odebatch, tspan1, [lBnd, n0(lBnd), 0], [], h, gr) ; 
    xmat(tspan1,k) = x(:,3) ;
 end

The first version of the code was working (I guess) in your first situation, because q used as a loop index (1:20) could also be used as a column index.