Quantum Generative Adversarial Networks in a Continuous-Variable Architecture to Simulate High Energy Physics Detectors
arXiv:2101.11132
Abstract
Deep Neural Networks (DNNs) come into the limelight in High Energy Physics (HEP) in order to manipulate the increasing amount of data encountered in the next generation of accelerators. Recently, the HEP community has suggested Generative Adversarial Networks (GANs) to replace traditional time-consuming Geant4 simulations based on the Monte Carlo method. In parallel with advances in deep learning, intriguing studies have been conducted in the last decade on quantum computing, including the Quantum GAN model suggested by IBM. However, this model is limited in learning a probability distribution over discrete variables, while we initially aim to reproduce a distribution over continuous variables in HEP. We introduce and analyze a new prototype of quantum GAN (qGAN) employed in continuous-variable (CV) quantum computing, which encodes quantum information in a continuous physical observable. Two CV qGAN models with a quantum and a classical discriminator have been tested to reproduce calorimeter outputs in a reduced size, and their advantages and limitations are discussed.
References in corpus (7)
- NIPS 2016 Tutorial: Generative Adversarial Networks
- Quantum computation over continuous variables
- Quantum generative adversarial learning
- CaloGAN: Simulating 3D High Energy Particle Showers in Multi-Layer Electromagnetic Calorimeters with Generative Adversarial Networks
- Strawberry Fields: A Software Platform for Photonic Quantum Computing
- PennyLane: Automatic differentiation of hybrid quantum-classical computations
- Mixing nonclassical pure states in a linear-optical network almost always generates modal entanglement