The paper Etalumis: Bringing Probabilistic Programming to Scientific Simulators is a best paper finalist at Supercomputing 2019, which will take place Nov 17-22, 2019 in Denver. Prof. Frank Wood of UBC Computer Science is the senior author and his former postdoc at Oxford Güneş Baydin is the first author.
Etalumis: Bringing Probabilistic Programming to Scientific Simulators at Scale
Atılım Güneş Baydin, Lei Shao, Wahid Bhimji, Lukas Heinrich, Lawrence Meadows, Jialin Liu, Andreas Munk, Saeid Naderiparizi, Bradley Gram-Hansen, Gilles Louppe, Mingfei Ma, Xiaohui Zhao, Philip Torr, Victor Lee, Kyle Cranmer, Prabhat, Frank Wood. Proc. Supercomputing 2019.
This work constitutes the largest-scale deployment of probablistic programming to date, in collaboration with Intel and Lawrence Berkeley National Laboratories. Their probablistic programming system PyProb is used to control multi-million line C++ Large Hadron Collider codebase, to do science in a novel way. The experiments in this paper were run at NERSC, using 32K cores on one of the largest supercomputers in the world.