Variational Autoencoders for Jet Simulation
arXiv:2009.04842
Abstract
We introduce a novel variational autoencoder (VAE) architecture that can generate realistic and diverse high energy physics events. The model we propose utilizes several techniques from VAE literature in order to simulate high fidelity jet images. In addition to demonstrating the model's ability to produce high fidelity jet images through various assessments, we also demonstrate its ability to control the events it generates from the latent space. This can be potentially useful for other tasks such as jet tagging, where we can test how well jet taggers can classify signal from background for events generated by the VAE. We test this idea by seeing the signal efficiency vs background rejection for different types of jet images produced by our model. We compare our VAE with generative adversarial networks (GAN) in several ways, most notably in speed. The architecture we propose is ultimately a fast, stable, and easy-to-train deep generative model that demonstrates the potential of VAEs in simulating high energy physics events.
30 pages, 25 figures, 4 tables
References in corpus (26)
- A Brief Introduction to PYTHIA 8.1
- Neural Discrete Representation Learning
- Identifying Boosted Objects with N-subjettiness
- CaloGAN: Simulating 3D High Energy Particle Showers in Multi-Layer Electromagnetic Calorimeters with Generative Adversarial Networks
- NVAE: A Deep Hierarchical Variational Autoencoder
- Variational Autoencoder for Deep Learning of Images, Labels and Captions
- Classification without labels: Learning from mixed samples in high energy physics
- Accelerating Science with Generative Adversarial Networks: An Application to 3D Particle Showers in Multi-Layer Calorimeters
- Searching for New Physics with Deep Autoencoders
- Anomaly Detection for Resonant New Physics with Machine Learning
- How to GAN LHC Events
- IntroVAE: Introspective Variational Autoencoders for Photographic Image Synthesis
- Getting High: High Fidelity Simulation of High Granularity Calorimeters with High Speed
- The Metric Space of Collider Events
- Uncovering latent jet substructure
- Learning to Classify from Impure Samples with High-Dimensional Data
- Regressive and generative neural networks for scalar field theory
- Tag N' Train: A Technique to Train Improved Classifiers on Unlabeled Data
- On the Topic of Jets: Disentangling Quarks and Gluons at Colliders
- Fast and accurate simulation of particle detectors using generative adversarial networks
- Controlling Physical Attributes in GAN-Accelerated Simulation of Electromagnetic Calorimeters
- Lund jet images from generative and cycle-consistent adversarial networks
- Image-based model parameter optimization using Model-Assisted Generative Adversarial Networks
- GANs for generating EFT models
- LHC analysis-specific datasets with Generative Adversarial Networks
- Information bottleneck through variational glasses