Talk Title: Inference in Causal Probabilistic Programs via Non-Standard Execution

This talk will introduce mechanisms for inference in the emerging paradigm of causal probabilistic programming, and simulation-based inference in general. First, Dr. Tavares will introduce ChiRho, a new system for automated and semi-automated causal inference, probabilistic inference, and arbitrary combinations thereof, within complex simulation models. Second, he will introduce parametric inversion, a theoretical and algorithmic framework for inverting non-injective, and hence non-invertible, programs. Parametric inversion executes programs in reverse, and in doing so points towards a new direction of inference methods in simulation models. Finally, he will discuss non-standard program execution as the common thread between these projects as well as a larger trend of inference methods that inspect and manipulate the structure of simulation models for inference. Non-standard interpretation changes the meaning of simulation models, and in doing so, acts as an advanced interface to new kinds of inference methods.