Event Details

With a rate of approximately 1 billion proton-proton collisions per second at an energy of 13 TeV, data sets from high energy physics collected at the Large Hadron Collider (LHC) are ideal for the application of machine learning. As new particles are created and detected, they produce high-dimensional, multi-modal streams of information that can be cast as sequential, imagebased, causal learning tasks. This talk will explore applications of computer vision techniques to improve generative and discriminative capabilities at the LHC. Specifically, it will outline the methodologies used in a recent work which introduced a deep generative model to enable high-fidelity, fast, detector simulation with preliminary speed-up factors of up to 100,000x. Although there are still open challenges, this work represents a significant stepping stone toward a full neural network-based simulator that could save significant computing time and enable many analyses at the LHC and beyond.