GENERATIVE MODELS USING CONTINUOUS VARIABLE QUANTUM COMPUTING

Monte Carlo methods have been used simulate the detectors of Large Hadron Collider (LHC) at CERN. But the process of simulating the detectors using Monte Carlo methods is computationally expensive and time consuming. One of the detectors of LHC, namely Electromagnetic Calorimeter which detects energies of particles, outputs continuous data. In this work we explore the inherent continuous nature of the Continuous Variable Architecture and try to find out whether or not we can take advantage of that and can simulate the detector efficiently using Generative Models with the current Quantum Software and Hardware technologies at our hand. Here, we implement two different models, a Quantum Generative Adversarial Network and a Born Machine in Continuous Variable Architecture using Pennylane as a Quantum Programming framework and Strawberryfields as a Backend by Xanadu. We analyze the results and determine that the implemented models perform well in learning certain types of probability distributions and not so well for the others. We discuss about the reasons this behaviour of the models and what can be the potential ways to tackle problem in future.