Automatic Code Generation for Statistical Models with Augmentation and Collapsing
thesisposted on 19.01.2021, 04:37 by SACHITH HASARANGA SENEVIRATNE
This thesis presents a system for automatically generating code for a set of Bayesian statistical models. The overall contributions include a system for performing pattern matching and symbolic computations, a schema system of efficiently managing statistical operations in a reusable manner. Additionally, the overall system for generating efficient code is also presented, alongside the general architecture that enables this process.