Accelerating Science with Generative Adversarial Networks: An Application to 3D Particle Showers in Multilayer Calorimeters

Abstract

Physicists at the Large Hadron Collider (LHC) rely on detailed simulations of particle collisions to build expectations of what experimental data may look like under different theoretical modeling assumptions. Petabytes of simulated data are needed to develop analysis techniques, though they are expensive to generate using existing algorithms and computing resources. The modeling of detectors and the precise description of particle cascades as they interact with the material in the calorimeter are the most computationally demanding steps in the simulation pipeline. We therefore introduce a deep neural network-based generative model to enable high-fidelity, fast, electromagnetic calorimeter simulation. There are still challenges for achieving precision across the entire phase space, but our current solution can reproduce a variety of particle shower properties while achieving speedup factors of up to 100 000×. This opens the door to a new era of fast simulation that could save significant computing time and disk space, while extending the reach of physics searches and precision measurements at the LHC and beyond.

Machine learning templates for QCD factorization in the search for physics beyond the standard model

“…A challenge for assessing previous GAN applications in HEP is that they have been designed to model high-dimensional feature spaces that are difficult to visualize and study [11][12][13][14][15][16][17][18][19][20][21]. The mass distribution example presented here provides a concrete testing ground to study GAN approaches where quantitative agreement can be studied and achieved using existing techniques.…”

“…2). GANs have also been studied in HEP and show great promise for accelerating simulations [11][12][13][14][15][16][17][18][19][20][21] and may also be useful for other tasks such as sampling from the space of effective field theory models [22]. In the context of QCD factorization studied in this paper, the GAN will learn the probability distribution of the jet mass given the jet kinematics and any other useful information.…”

Supervised Deep Learning in High Energy Phenomenology: a Mini Review*

“…In [78] the feasibility of deep generation model in high-quality, fast and electromagnetic calorimeter simulation is further studied. It was found that although it is difficult to accurately simulate the whole phase space with deep generation model, this method can reproduce many simulation properties of the detector and can accelerate the simulation of calorimeter by 100,000 times.…”

Neural hierarchical models of ecological populations

“…For example, models of lake temperature that combine neural networks with loss functions consistent with known physical mechanisms perform better than physical models and neural networks applied alone (Karpatne et al 2017). Similarly, generative adversarial networks with loss functions that encourage mass-balance have expedited electromagnetic calorimeter data generation from the Large Hadron Collider (Paganini et al 2018;Radovic et al 2018). Convolutional neural networks also have been successfully deployed in population genetics to make inferences about introgression, recombination, selection, and population sizes (Flagel et al 2018).…”