Generative Adversarial Network based method for generation of synthetic image parameters for TACTIC γ-ray telescope

Very High Energy (VHE) γ and cosmic rays interact with the atmosphere and generate an Extensive Air Shower (EAS) of particles comprising mainly electron-positron pairs. These energetic secondary charged particles move down the atmosphere and upon de-excitation, the medium generates a flash of Cherenkov light. The subtle differences in the shape of Cherenkov flash images generated by γ/cosmic rays at the top of the atmosphere can be explored by the ground-based Imaging Atmospheric Cherenkov technique (IACT) telescopes. To characterize and calibrate such IACT telescopes, an extensive library of EAS for γ-rays and cosmic rays of various energies are required. However, the main difficulty in generating a simulation database is that the process of shower generation is computationally extensive. To overcome the challenge of creating a huge database of simulated shower images, we propose a method based on the generation of synthetic data by employing the unsupervised machine learning (ML)-based method, namely the Generative Adversarial Networks (GAN). The present study discusses the methodology for generating synthetic image parameters using GAN for the TACTIC telescope. A comparison between Monte Carlo simulation-generated image parameters for the TACTIC and GAN-generated parameters is presented. Results are also presented from the independent Random Forest-based ML classifier to identify similarities between the simulation and synthetic data species. The potential applications of GAN-generated data towards IACT data analysis for TACTIC are also discussed. [Display omitted] •The database generation for IACT telescopes is computationally intensive.•In astronomy, Generative machine learning models can be used for synthetic data generation.•A trained GAN network trained by MC data can produce Hillas parameters very swiftly.•Produced synthetic data are statistically indistinguishable from training data.•Apart from saving CPU hours, GAN interpolated data has other potential applications in IACT data analysis.