How to access/manipulate data from cell arrays?

Asked by Thomas Veith

Thomas Veith (view profile)

on 3 Jun 2019
Latest activity Commented on by Adam Danz

on 3 Jun 2019

Adam Danz (view profile)

So I have the code bleow which sovles an ODE 1000 times with random variables. The result is a 1000 by 1 cell array, the contents of which are all XX by 2 numeric arrays (column 1 solves x(1), column 2 solves x(2)). What I would like to produce is an XX by 1000 numeric array, where my 1000 columns would be comprised of all column 2 data from the numeric arrays nested in my cell array. Is this possible? I've looked through the Cell Arrays and Multilevel Indexing to Access Parts of Cells pages, and haven't been able to find what I'm looking for. Any help would be appreciated.
n = 1000;
result = cell(n,1);
for k=1:n
tspan=[1 7];A0=rand;P0=rand;g=rand;p=rand;B=rand;
[t,x] = ode45(@(t,x) [-g*x(1) + p*x(1); -x(1)*x(2)+ B*x(2)], tspan, [A0 P0]);
result{k} = x;
end

Adam Danz (view profile)

on 3 Jun 2019
Edited by Adam Danz

Adam Danz (view profile)

on 3 Jun 2019

This extracts column 2 data from all elements of the cell array and arranges it in a row within a cell array as you described.
C = cellfun(@(x)x(:,2),result,'UniformOutput', false)';
Here's how to put that into a padded array
% how much padding is needed per cell?
maxLen = max(cellfun(@(x)size(x,1),result)); %max length
padNum = maxLen - cellfun(@(x)size(x,1),result);
% get the col 2 vals and pad them
C = cellfun(@(x)x(:,2),result,'UniformOutput', false);
% Put them into a matrix

Show 1 older comment

Adam Danz (view profile)

on 3 Jun 2019
It sounds like you're describing a matrix with varying number of rows which isn't possible. However, it is possible to produce a padded array where NaNs (or any other value you want, 0s for example) can be used to fill spaces. I'll update my answer to show how.
Thomas Veith

Thomas Veith (view profile)

on 3 Jun 2019
Thank you! What if they were all the same size? I get that I could write a matrix M123=[C{1,1},C{1,2},C{1,3},...,C{1,1000}]; but that would be quite tedious.

Adam Danz (view profile)

on 3 Jun 2019
If they were all the same size, it would be as easy as
C = cellfun(@(x)x(:,2),result,'UniformOutput', false)';
cell2mat(C)

madhan ravi (view profile)

on 3 Jun 2019
Edited by madhan ravi

madhan ravi (view profile)

on 3 Jun 2019

result{k} = x(:,2);
To produce a numeric matrix you need to append with nan or zeros at the end.
M=max(cellfun('prodofsize', result));
Desired=cell2mat(cellfun(@(x) [x;zeros(M-numel(x),1)], result,'un', 0))

Thomas Veith

Thomas Veith (view profile)

on 3 Jun 2019
Thank you for your quick reply! That works very well, however, it's not exactly what I was looking for. That changes my result so that my cell array only includes the results from column 2, each selection of which represents a numerica array of varying size (41x1, 285x1, etc, etc). That is nice, and more managable than including both x solutions, but, what I was really hoping for was that I could produce a YYx1000 numeric array, so that column one would be the 41 entries, column two the 285 entries, etc, etc. Is this possible? Thanks in advance.