methods for approximate Bayesian inference, and gradient-based optimization for penalized maximum likelihood estimation.