[julia-users] Efficient convolution of sparse arrays?

['Antoine Messager' via julia-users]

Hello,

I have signals that are 12,654,646 seconds long but very sparse, there are 
only a few hundreds events. I would like to compute the cross-correlation 
of those signals in a fast manner. Up to now I do the following:

cross_cor=zeros(Int,120*60)
for t0 in sig1
    for t in sig2
        if t0<=t<t0+120*3600-1
            cross_cor[floor(Int,(t-t0)/60)+1]+=1
        end
    end
end

In sig1 and sig2, I have the instant of the events, which are thus array of 
size a few hundreds. This is thus much faster than using the crosscor 
<https://github.com/rened/Stats.jl> function of the Stats.jl package. I 
would like to know if there would exist any way to quicken this? Or in 
other word is there any efficient way of convolving two sparse arrays?

Many thanks,

Antoine