Speaker
Description
We present a specialised layer for generative modeling of LHC events with generative adversarial networks. We use Lorentz boosts, rotations, momentum and energy conservation to build a network cell generating a 2-body particle decay. This cell is stacked consecutively in order to model two staged decays, respecting the symmetries across the decay chain. We allow for modifications of the resulting four-vectors in order to model higher order and detector effects.
We give an evaluation of the generator quality in a higgs decay into two
Z bosons, further decaying into a muonpair each.
Significance
In the High-Luminosity LHC era, the requirement for Monte-Carlo simulation will outscale the available resources. Generative adversarial networks show a promising way to generate events at very low computational cost. First efforts have been done in this field but there are very few specialised architectures that incorporate the needed physisc symmetries.
Speaker time zone | Compatible with Europe |
---|