Getting High: High Fidelity Simulation of High Granularity Calorimeters with High Speed

Abstract

Accurate simulation of physical processes is crucial for the success of modern particle physics. However, simulating the development and interaction of particle showers with calorimeter detectors is a time consuming process and drives the computing needs of large experiments at the LHC and future colliders. Recently, generative machine learning models based on deep neural networks have shown promise in speeding up this task by several orders of magnitude. We investigate the use of a new architecture—the Bounded Information Bottleneck Autoencoder—for modelling electromagnetic showers in the central region of the Silicon-Tungsten calorimeter of the proposed International Large Detector. Combined with a novel second post-processing network, this approach achieves an accurate simulation of differential distributions including for the first time the shape of the minimum-ionizing-particle peak compared to a full Geant4 simulation for a high-granularity calorimeter with 27k simulated channels. The results are validated by comparing to established architectures. Our results further strengthen the case of using generative networks for fast simulation and demonstrate that physically relevant differential distributions can be described with high accuracy.

Hadrons, Better, Faster, Stronger

“…In Refs. [14,15] we showed a precise modelling of differential distributions over many orders of magnitude for electromagnetic showers. The present work extends this level of precision for the first time to the more challenging hadron-induced showers in a highly granular hadronic calorimeter.…”

“…The WGAN presented here is similar to the one used for photon shower generation in Ref. [14], however two architectural changes are applied in order to improve its generative performance for hadronic showers.…”

“…The BIB-AE presented here is a significant extension of the setup used for photon shower generation in Ref. [14]. We therefore first describe the baseline model followed by a detailed discussion of the extensions.…”

“…Furthermore, the loss function of the Post Processor was also refined. In addition to the Mean Squared Error (MSE) and Sorted Kernel Maximum Mean Discrepancy (SK-MMD) described in [14] we implemented a series of auxiliary loss terms.…”

“…Motivated by the large fraction of resources already consumed by calorimeter simulation [5], and the expected increase due to higher granularities and luminosities, the precise and fast simulation of calorimeters is a primary topic of research [2,[6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23]. 1 In the quest to simulate calorimeters with high accuracy, standard generative machine leanring methods -Generative Adversarial Networks (GANs) [26], Variational Autoencoders (VAEs) [27] and normalising flows [28][29][30][31] -have all been considered.…”