12-01, 11:10–11:25 (Europe/Amsterdam), Ernst-Curie
Julia has established itself as the programming language that sits in the intersection
between very fast languages, like C and Fortran, and high abstraction
programming languages, like Python and Matlab. It has been argued
that this solves the 2 language problem where a prototype is programmed
in a high abstraction language and later needs to be re-implemented in a
fast language. In this talk, we will argue that, while Julia does solve the 2
language problem, it also creates the 1.5 language problem where there is a
huge difference between working Julia code and very efficient Julia code. We
will showcase a medium-sized problem to highlight this discrepancy. Finally,
we want to start a discussion in the community on how to solve this problem
by targeting the education of Julia novices towards performance at scale.
A short talk (either 10 or 20 minutes) to present our own learning experience with respect to creating a medium sized dynamics model and then making it fast. We want to start a discussion about how to discuss performance aspects of the programming language and when to present these details. In particular, we believe that memory management is something that needs to be presented earlier and discussed more frequently, at least for experienced programmers for whom Julia is not their first programming language.
Michael Tiemann obtained a PhD from the University of Tübingen where he was working on a machine learning perspective of differential equation solvers. In 2017, he has joined the Bosch Center for Artificial Intelligence in Renningen, Germany. He is working on probabilistic inference, dynamics modeling and, in particular, combining data-based and first-principles modeling approaches. He has been eagerly waiting for a chance to switch to Julia since 2018, but the chance only arrived in Spring 2022.
A research scientist at Robert Bosch GmbH working on Active Learning and Hybrid Modelling with a particular interest in solutions involving Julia.