A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics

Tobias Fritz

We develop Markov categories as a framework for synthetic probability and statistics, following work of Golubtsov as well as Cho and Jacobs. This means that we treat the following…