Towards formal verification of Bayesian inference in probabilistic programming via guaranteed bounds

Thesis

Probabilistic models are an indispensable tool in many scientific fields, from the social and medical sciences to physics and machine learning. In probabilistic programming, such models are specified as computer programs: a flexible yet precise representation that allows for automated analysis. Such probabilistic programs can express both randomized algorithms and statistical or machine learning models. This thesis focuses on the Bayesian framework for reasoning and learning under uncertai...

Expand abstract

Authors: Zaiser, F

• DOI: 10.5287/ora-mnznara4o
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: probabilistic programming; guaranteed bounds; quantitative analysis; Bayesian statistics; program analysis; formal methods; programming languages; formal verification; probabilistic inference
• Subjects: Formal methods (Computer science); Probabilistic Inference; Bayesian Statistics; Probabilistic Programming; Programming languages (Electronic computers)