Likelihood Free Inference

Likelihood-free inference (LFI) tackles the challenge of performing statistical inference on models with intractable likelihood functions, common in complex simulations across diverse scientific fields. Current research focuses on developing efficient and accurate LFI methods using neural networks, particularly normalizing flows and generative models, often coupled with algorithms like Robust Optimisation Monte Carlo (ROMC) for improved sample efficiency and parallelization. These advancements enable robust parameter estimation and uncertainty quantification in scenarios ranging from gravitational wave detection to brain imaging analysis, offering powerful tools for analyzing complex systems where traditional statistical methods are insufficient.