New technique boosts accuracy and efficiency in probabilistic programming

Researchers at KTH have developed an automated technique to improve the efficiency and accuracy of statistical inference in probabilistic programming languages (PPLs). The technique automatically determines checkpoint locations in probabilistic programs, eliminating the need for manual identification by developers. This automation streamlines the process and reduces errors in statistical inference, benefiting various research fields, including phylogenetics, computer vision, topic modeling, and cognitive science. The technique has received recognition and won the EAPLS Best Paper Award at the European joint conferences on theory and practice of software (ETAPS) in 2023.