For the lack of an exact mathematical theory, computer simulations are currently the most accurate tool to study the matter and energy distribution in the Universe, the so-called ""large scale structure"". Gathering the results of interest by running such simulations is, however, both very expensive and requires specialized infrastructure for high performance computing. As a consequence, tasks that rely on large sets of such results are rendered infeasible unless the simulations can be (partly) replaced by much faster but still highly accurate (machine learning) tools, in cosmology referred to as ""emulators"".  
In this talk I will present the cosmological emulator I have developed over the past years, emphasizing the challenges we faced during the construction process.

Download the slides for this talk.Download ( PDF, 71928.46 MB)