New angles on fast calorimeter shower simulation

Abstract

The demands placed on computational resources by the simulation requirements of high energy physics experiments motivate the development of novel simulation tools. Machine learning based generative models offer a solution that is both fast and accurate. In this work we extend the Bounded Information Bottleneck Autoencoder (BIB-AE) architecture, designed for the simulation of particle showers in highly granular calorimeters, in two key directions. First, we generalise the model to a multi-parameter conditioning scenario, while retaining a high degree of physics fidelity. In a second step, we perform a detailed study of the effect of applying a state-of-the-art particle flow-based reconstruction procedure to the generated showers. We demonstrate that the performance of the model remains high after reconstruction. These results are an important step towards creating a more general simulation tool, where maintaining physics performance after reconstruction is the ultimate target.

CaloClouds: fast geometry-independent highly-granular calorimeter simulation

“…Unlike previous image-based fast simulation models such as ref. [16,18,21], the computational cost of the PointWise Net scales O (𝑁) with the number of points and hence with the energy 𝐸 just like Geant4. Therefore, to improve overall training and sampling speed, we batch together events with a similar number of points.…”

“…The concept of generative calorimeter simulation is well-proven at this point. Its underlying mathematical viability has been demonstrated [4,5] and direct applications have been shown using generative adversarial networks [6][7][8][9][10][11][12][13][14][15], autoencoder-variants [16][17][18], normalizing flows [19][20][21] and diffusion models [22][23][24]. They have been successfully deployed e.g.…”

CaloScore v2: single-shot calorimeter shower simulation with diffusion models

“…The application of deep generative models to calorimeter simulation began with CaloGAN [1,2]. Since that time, Generative Adversarial Networks (GANs) [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17], Variational Autoencoders [15,16,[18][19][20][21], Normalizing Flows (NFs) [22][23][24][25][26][27][28][29], and Diffusion Models [30][31][32][33] have been applied to this problem. This research entered a precision era with the first NF application (CaloFlow) [23], which showed that even a post-hoc classifier had difficulty distinguishing physics from machine learning simulators.…”

CaloClouds II: ultra-fast geometry-independent highly-granular calorimeter simulation

“…Detector simulations, such as the simulation of the sensor response in highly granular calorimeters, can be augmented or sped up by employing modern generative machine learning methods [5][6][7][8]. Recent studies have explored the simulation of calorimeter showers with various generative models such as generative adversarial networks (GANs) [5,[9][10][11][12][13][14][15][16][17][18][19], autoencoders and their variants [20][21][22][23][24][25], and normalizing flows [26][27][28][29][30][31][32][33]. Additionally, diffusion models [34][35][36][37][38], also referred to as score-based generative models, have been shown to provide very high fidelity on calorimeter data [39][40][41][42][43].…”