Emerging Computational Techniques for Jet Physics

Jets are the most copiously produced objects at the LHC and the subject of intense experimental and theoretical study. Improvements to our understanding and treatment of jets can have significant impact on the physics program of the LHC; however, various computational bottlenecks appear in this quest. Below we will discuss a few areas where such computational bottlenecks appear and identify emerging computational techniques that may be able to address them. We hope that this may challenge some assumptions about the computational demands of simulation, reconstruction, and analysis of LHC data when jets are involved. Probabilistic programming systems use the simulator directly during inference. They are integrated with the control flow of the event generator though they require a tractable joint likelihood. Thus, they offer both a natural way to improve the simulation efficiency, as well as the possibility of doing systematic fits to high-dimensional data by a more direct implementation of backpropagation.