Causal Theories: A Categorical Perspective on Bayesian Networks — Brendan Fong (2013). arXiv:1301.6201 (v1, PDF).
Master’s dissertation proposing causal theories — symmetric monoidal categories whose objects are variables and whose morphisms deduce information about one variable from another — as an algebraic framework for Bayesian networks and causal reasoning, with graphical representations matching information flow.
Sources: the paper, arXiv:1301.6201v1, checked against the arXiv listing. Index: Papers.
Key definitions and results
- Causal theories; Bayesian networks as string diagrams (Theorem 4.5, as cited by AutoBayes)