Event generation with normalizing flows

We present a novel integrator based on normalizing flows which can be used to improve the unweighting efficiency of Monte Carlo event generators for collider physics simulations. In contrast to machine learning approaches based on surrogate models, our method generates the correct result even if the underlying neural networks are not optimally trained. We exemplify the new strategy using the example of Drell-Yan type processes at the LHC, both at leading and partially at next-to-leading order QCD.

Figure 1

Structure of a coupling layer. $m$ is the output of a neural network and defines the coupling transform, $C$, that will be applied to $x_B$.

Figure 2

Projections of the sampled parameters and color coded acceptances. The plot on the left suggest a high learning rate, coupled with a large number of epochs to be beneficial. The plot on the right suggests a strong preference for a small number of bins. The best performing configuration is indicated with a star.