How can I isolate data from a matrix without a for loop?

Dear all

given the matrix

M = rand(1000,10);

pre_indices = [ 40 60 70]

post_indices = [45 65 75]

I would like to isolate the following chunks from M

for i = 1:length(pre_indices)

M(pre_indices(i):post_indices(i),:)

end

could it be done without the loop? all chunks have the same dimensions

best regards MF

[Guillaume]

It's debatable whether using accumarray or arrayfun and co. can be considered to be without a loop. Yes, it's not a for loop but all these functions mean more or less the same: "loop over the data and do the same thing to each element".

In your case, the loop may be the clearest and fastest way to obtain what you want.

Answer 1

[Andrei Bobrov]

M = (1:100)'*ones(1,5);
pre_indices = [ 40 60 70];
post_indices = [45 65 75];

n = (1:size(M,1))';
l = (n >= pre_indices(:)' & n <= post_indices(:)')*(1:numel(pre_indices))';

out = accumarray(l(l>0),n(l>0),[],@(x){M(x,:)});

or

n = (1:size(M,1))';
l = (n >= pre_indices(:)' & n <= post_indices(:)').*n;
out = M(l(l>0),:);

or

n = (1:size(M,1))';
l = any(n >= pre_indices(:)' & n <= post_indices(:)',2);
out = M(l,:);

if you have older version of MATLAB (< R2016b)

n = (1:size(M,1))';
k = bsxfun(@ge,n,pre_indices(:)') & bsxfun(@le,n,post_indices(:)');

l = k*(1:numel(pre_indices))';
out = accumarray(l(l>0),n(l>0),[],@(x){M(x,:)});

or

l = bsxfun(@times,k,n);
out = M(l(l>0),:);

or

out = M(any(k,2),:);

Answer 2

[Guillaume]

arrayfun(@(si, ei) M(si:ei, :), pre_indices, post_indices, 'UniformOutput', false)

This is most likely slower than the explicit loop, and for all intents and purposes is a loop.