Using accumarray to create histograms from random integers

Hi,

I'd like to create N histograms, each from a dataset of n_vals random integers from 1 to h_length. I can demonstrate my goal for N=1 using accumarray:

N=1;
h_length=10000;
n_vals=900;
%create n_vals random integers between one and h_length
inds=floor((h_length)*rand(n_vals,N)+1);
rand_hists=accumarray(inds,1,[h_length,1]);
%this plot demonstrates the output.
figure;plot(rand_hists)

However, I'm not sure how to do this for an N larger than 1. Ultimately, I want N~=2000. I'm pretty sure I will have a memory limitation and have to do some looping, but I would like to vectorize and do it as efficiently as possible.

Any advice is greatly appreciated.

Thanks, Luke

Comment on this

Answer 1

[Fangjun Jiang]

Check to see if you have these function.

randi()
hist()

[Luke S]

I have both! randi() I didn't know about before, and it seems to save a little bit of time over floor(randn()).

Using histc() can operate on an Nxh_length array of inds as I wanted, but it actually seems o be ending up slower than looping over accumarray N times.

Answer 2

[Peter Perkins]

Luke, I'm not exactly sure what you're asking, but as I interpret it, you want a single call to accumarray to create N sets of counts. You may find that this is no faster than a loop around accumarray, but the following code is what I think you're looking for. It passes accumarray 2-D indices (which can be done in two different ways, here I used separate column vectors), where the second index is just an "indicator variable that says which "set" of counts a given row should be put in.

I've replaced your call to rand with randi, and I'm not sure what you're plot was supposed to be doing, but I more or less kept it as is.

N=3;
h_length=10000;
n_vals=900;
%create n_vals random integers between one and h_length
iinds = randi(h_length,n_vals,N);
jinds = repmat(1:N,n_vals,1);

rand_hists = accumarray({iinds(:) jinds(:)},1,[h_length,N]);

%this plot demonstrates the output.
for i = 1:N
    subplot(N,1,i), plot(rand_hists(:,i),'.')
end