Poisson-beta Distribution

Recently I’m developing a neural network that uses probability distributions to fit chemical reaction processes. Biologically, the Poisson-beta distribution is what we are interested in. But its original formula is probably an integral that cannot be calculated. Note that an approximation can be made using Gaussian-Jacobian quadrature. In fact, scipy.special.roots_jacobi can be used in scipy to calculate Gauss-Jacobi orthogonal sampling points and weights. I’m curious if we can implement a Poisson-beta Distribution class for torch.distributions based on this :slight_smile: