How to transpose cell arrays within a larger cell array

Hi, I was wondering how to use cellfunc to find and transpose cell arrays that are row orientated within another cell array. I can do it using a loop but I was curious how to use cell function.

Code is below, I think there is a flaw in the method because I am using a loop mentality using seperate variables.

The dynamics file is a {1x3} cell and embedded within this are another set of {3x1} cells with uniform matrices. The third cell is the one that is row and ebThe 3rd cell is the one I can change, but I still think that on a larger scale my method will cause probelms.

If someone could have a quick it would be appreciated.

Thanks

load('dynamics.mat')

%find rows in cell array
idx = find(cellfun(@isrow, dyn));
trans_dyn = (cellfun(@transpose, dyn(idx), 'UniformOutput', false));
dyn(idx) = trans_dyn;

Answer 1

[Jan]

"The dynamics file is a {1x3} cell and embedded within this are another set of {3x1} cells with uniform matrices." - What are "uniform matrices?

"I still think that on a larger scale my method will cause probelms." - Why do you assume this?

idx      = cellfun(@isrow, dyn);   % Omit the FIND for faster logical indexing
dyn(idx) = cellfun(@transpose, dyn(idx), 'UniformOutput', false);

Or easier in 1 step:

dyn = cellfun(@(c) c(:), dyn, 'UniformOutput', false)

All these methods are slow.

Data = load('dynamics.mat');
dyn0 = repmat(Data.dyn, 1, 1e5);  % A larger set

dyn = dyn0;
tic;
idx = find(cellfun(@isrow, dyn));
trans_dyn = (cellfun(@transpose, dyn(idx), 'UniformOutput', false));
dyn(idx) = trans_dyn;
toc  % Elapsed time is 1.827274 seconds.

dyn = dyn0;
tic;
idx      = cellfun(@isrow, dyn);   % Omit the FIND for faster logical indexing
dyn(idx) = cellfun(@transpose, dyn(idx), 'UniformOutput', false);
toc  % Elapsed time is 1.794807 seconds.

dyn = dyn0;
tic;
dyn = cellfun(@(c) c(:), dyn, 'UniformOutput', false);
toc  % Elapsed time is 3.947908 seconds.

dyn = dyn0;
tic;
dyn = cellfun(@(c) reshape(c, [], 1), dyn, 'UniformOutput', false);
toc  % Elapsed time is 2.599826 seconds.

dyn = dyn0;
tic;
for k = 1:numel(dyn)
    dyn{k} = dyn{k}(:);
end
toc  % Elapsed time is 1.523816 seconds.

dyn = dyn0;
tic;
for k = 1:numel(dyn)
    dyn{k} = reshape(dyn{k}, [], 1);
end
toc  % Elapsed time is 0.515385 seconds.

dyn = dyn0;
tic;
for k = 1:numel(dyn)
    if isrow(dyn{k})
        dyn{k} = dyn{k}(:);
    end
end
toc  % Elapsed time is 0.744554 seconds.

dyn = dyn0;
tic;
for k = 1:numel(dyn)
    if isrow(dyn{k})
        dyn{k} = reshape(dyn{k}, [], 1);
    end
end
toc  % Elapsed time is 0.433867 seconds.

dyn = dyn0;
tic;
for k = 1:numel(dyn)
    if isrow(dyn{k})
        dyn{k} = dyn{k}.';
    end
end
toc  % Elapsed time is 0.392983 seconds.

Prefer loops. cellfun is nice, but slower than a loop.

[Richard Rees]

Hi Jan, my code creates consistent sized files of parameters extracted from datafiles that are based on a known number of Node and Gau points from which they are derived. The dynamics is on the of the variables I create that is based on derviations of this data.

I am very rusty with cell functions, prefer loops for security.

Many thanks

Richard