I am working on modeling species invasion using an integro differential equation framework,
N _{r,t+1} = \int_{0}^{\infty} \int_{0}^{2\pi} k(|r-r'|) f(N _{r',t}) dr'd\theta,
where N _{r,t} is the population at location r at time t and f(N _{r,t}) is the growth function of the species. A naive approach to solve this integral is by running a for loop which has time complexity \mathcal{O}(n^2).
Alternatively, the convolution can be solved using 2D FFT
N _{r,t+1} = \text{ifft2}(\text{fft2}(N _{r,t})*\text{fft2}(k(|r-r'|))), for which complexity scales as \mathcal{O}(n\log{n}).
I can easily compute the convolution in stan using fft and ifft functions. But since my simulations are computationally demanding, I was wondering if the stan team is working on providing 1D and 2D convolution functions. This could help me save a lot of time since this is the most computationally intensive part of my simulations.