Help for vectorization??

incI=zeros(length(featInd),length(tempInd));

for f = 1:length(featInd)  
    incI(f,:) = tempInd + featInd(f);
end

for k=1:length(featInd)
    h([incI(k,:)]') = h([incI(k,:)]')+ 1;
end

[jgg]

You need to provide more information. What is h? What is featInd? What is tempInd? How are they structured?

[Parag Patil]

featInd=[1,24636] row vector

tempInd=[1,12332] row vector

h=[14593760,1] coloumn vector

Answer 1

[Honglei Chen]

Here is an example you can try

featInd = 1:24636;
tempInd = 1:12332;
incI = bsxfun(@plus,featInd(:),tempInd);

As to the second loop, I could be wrong but that looks just like a histogram for me, so you can probably do something like

[incU,~,incInd] = unique(incI(:),'stable');
h = zeros(14593760,1);
h(incU) = h(incU)+accumarray(incInd,ones(numel(incInd),1));

[Honglei Chen]

Updated the code with correct dimension, although the dimension of h may depend on what you have in featInd and tempInd

[Parag Patil]

Thanks Honglei Chen tried your approach for second loop. Works for smaller values of tempInd & featInd. For large values unique function gives out of memory error when executed on GPU.