The Geometry of Bayesian Programming
We give a geometry of interaction model for a typed lambda-calculus with operators for sampling and soft conditioning. The calculus can as such be seen as a paradigmatic calculus for higher-order bayesian programming languages, like ANGLICAN or CHURCH. We present the model in two flavors, the former corresponding to an idealized, thus not executable, learning algorithm, the latter allowing to see the learning algorithm as part of the of the environment.
Tue 15 Jan
|11:00 - 11:30|
|11:30 - 12:00|
Eli SenneshNortheastern University, Adam ŚcibiorUniversity of Cambridge and MPI Tuebingen, Hao WuNortheastern University, Jan-Willem van de MeentNortheastern UniversityFile Attached
|12:00 - 12:30|
David TolpinPUB+Media Attached