Optimization using pathwise algorithmic derivatives of electromagnetic shower simulations

Max Aehle, Mihály Novák, Vassil Vassilev, Nicolas R. Gauger, Lukas Heinrich, Michael Kagan, David J. Lange

Abstract: Highlights•We have successfully applied algorithmic differentiation to a stochastic particle simulation.•The entire electromagnetic shower simulation is “AD-friendly”, except for the implementation of multiple scattering.•The resulting gradient estimator can be useful for optimization in high- energy physics.