Accelerating Binary Neutron Star Merger Simulations with Artificial Neural Networks

Artificial neural networks (ANNs) are fast becoming an important element in the toolkit we use to tackle complex computational problems, with GPUs making up the majority of available horsepower on many of the latest generation of computing clusters. In this talk I will discuss how ANNs can be used to accelerate a modern general-realtivistic radiation magneto-hydrodynamics code aimed at exploring binary neutron star (BNS) mergers. In particular I will give a brief description of ANNs and how they work, explore where ANNs may be useful in the context of simulating BNS mergers, and demonstrate that ANNs are able to increase the speed of even our CPU based code.