can i velocize it? (if possible to vectorize)

[r,c]=size(DD);

z=find(RP_bin);

filtro=zeros(numel(z),c);
period=3; %value=from 1 to numel(z)
tic
    for i=1:c
        for ii=period+1:numel(z)
           idx=z(ii);
           idx2=max(1,MinTrade_Tab(ii,i));
           s=DD(idx2:idx,i);
           n=any(ismember(s,0));  
           filtro(ii,i)=n;
        end
    end
    toc

[dpb]

You would stand a far better chance if you explained the objective here and what the input data are instead of expecting folks to ferret it out on their own...

[shamal]

I posted the complete data I was convinced that not all this information was needed to speed up an algorithm. DD is the time series and I note if there is a value==0 between the 2 indices examined (idx and idx2)

MinTrade_Tab contains index

Answer 1

[Bruno Luong]

On my computer the acceleratoion of Method V2 is more impressive

load('matlab_DD.mat')
load('matlab_minTrades.mat')
load('matlab_RpBin.mat')

[r,c]=size(DD);

z=find(RP_bin);
period = 0 % not provided by OP

period = 0

tic
filtro=zeros(numel(z),c);
for i=1:c
    for ii=period+1:numel(z)
        idx=z(ii);
        idx2=max(1,MinTrade_Tab(ii,i));
        s=DD(idx2:idx,i);
        n=any(ismember(s,0));
        filtro(ii,i)=n;
    end
end
toc

Elapsed time is 0.027284 seconds.

% Method V2
tic
[m,n] = size(DD);
k = [-Inf; find(DD==0); Inf];
start = MinTrade_Tab(period+1:numel(z),:);
start = max(start,1);
stop = z(period+1:end);
kstop = stop + m * (0:n-1);
kstart = start + m * (0:n-1);
l1 = discretize(kstart,k,'IncludedEdge','right');
l2 = discretize(kstop,k,'IncludedEdge','left');
filtro2 = l1+1<=l2;
filtro2 = [zeros(period,c); filtro2];
toc

Elapsed time is 0.012275 seconds.

close all
ax1=subplot(1,2,1);
imagesc(filtro)
ax2=subplot(1,2,2);
imagesc(filtro2)
linkaxes([ax1 ax2])

% Check
isequal(filtro, filtro2)

ans = logical
   1