OASIS: Optimal Analysis-Specific Importance Sampling for event generation

Abstract

We propose a technique called Optimal Analysis-Specific Importance Sampling (OASIS) to reduce the number of simulated events required for a high-energy experimental analysis to reach a target sensitivity. We provide recipes to obtain the optimal sampling distributions which preferentially focus the event generation on the regions of phase space with high utility to the experimental analyses. OASIS leads to a conservation of resources at all stages of the Monte Carlo pipeline, including full-detector simulation, and is complementary to approaches which seek to speed-up the simulation pipeline.

Challenges in Monte Carlo Event Generator Software for High-Luminosity LHC

“…[105]. A completely different optimization strategy [116] for the sampling algorithm has also been recently proposed, where the goal is that of giving priority to populating the regions of phase space which are most sensitive to the presence of a signal or to the value of a parameter.…”

Speeding up MadGraph5_aMC@NLO

“…Although not shown in this table, roughly half of this computation (19% of the total) is spent evaluating simply scalar products (the contraction of Eqs. (20)(21)(22) with the remaining wave function) which strongly limit the hope of further optimisation.…”

“…In the context of helicity recycling a new type of helicity routine has been introduced (Eqs. (20)(21)(22)). Contrary to the other types of routines, MG5aMC does not know at generation time which of those functions will be effectively used.…”

“…Third, one can try to use more efficiently the various types of hardware (e.g. GPU, MPI, vectorization) [15][16][17][18] and finally one can play with the weights of the sample to optimise/re-use information [19][20][21][22][23][24].…”

Accelerating Monte Carlo event generation -- rejection sampling using neural network event-weight estimates

“…The use of MCMC techniques for exploring high-dimensional phase spaces has been studied in [35]. In [36] the application of analysis-specific optimal sampling distributions was proposed, similar to methods of biasing event generation, e.g. to oversample tails of physical distributions [37].…”