Integrating Stan with Elixir for Advanced Probabilistic Modeling

Shubham Gupta introduces the audience to probabilistic modeling using Elixir and Stan in his talk at ElixirConf EU 2024. He starts by explaining the fundamentals of probabilistic programming through an example from the game show 'Let's Make a Deal'. Shubham then delves into the concepts of Bayesian inference, describing its components like priors, likelihoods, and posteriors, and explains how these are used to compute probabilities based on observed data. He also covers methods to simplify Bayesian inference, such as using conjugate priors and sampling techniques, including Metropolis-Hastings and Hamiltonian Monte Carlo (HMC) sampling. Shubham introduces ExStan, a lightweight SDK for integrating Stan with Elixir, enabling advanced statistical modeling and probabilistic programming. He demonstrates this integration with case studies—including a simple linear regression using the Iris dataset and a more complex model for predicting outcomes of football matches during the FIFA World Cup. Finally, he discusses the limitations and future development plans for ExStan.