Have you ever seen hierarchical probabilistic models computed in milliseconds? Now it is possible, on a low-cost desktop machine, from a dynamic and interactive REPL environment in Clojure. This talk presents Bayadera, an opinionated Bayesian statistical library backed by a high-speed massively parallel MCMC engine that runs on the GPU. Bayadera has been designed with programmers in mind; it is optimized not only for dynamic, interactive model-building and experimentation, but also to provide fast performance in production. There is even more; Bayadera has been written in Clojure, which is a pragmatic modern Lisp dialect that runs on Java Virtual Machine, and plays well with the Java ecosystem. The talk will introduce interactive development in Clojure, an infrastructure for high-performance numerical computing provided by Uncomplicate libraries, and give a walk-through of real model building and fitting from one of the later chapters of the Doing Bayesian Data Analysis.
Dragan Djuric is a professor at the Department of Software Engineering, FON, University of Belgrade, Serbia. He passionately uses Clojure as a primary language since 2009, and teaches Clojure-based courses at the university since 2010. He published his Clojure-based research in leading scientific journals, but does not skip contributiong to the community through open-source Clojure projects (www.uncomplicate.org). His main interests are in the area of software engineering and intelligent systems, but programming in Clojure is the activity he enjoys the most. When he is not working in Emacs, he likes doing his daily dose of long-distance running, gym, and Cuban salsa dancing. He also has a black belt in Taekwondo, but is a very calm and friendly person :)