Supernova Light Curves Approximation based on Neural Network Models

Photometric data-driven classification of supernovae becomes a challenge due to the appearance real-time processing big data in astronomy. Recent studies have demonstrated superior quality solutions based on various machine learning models. These models learn classify supernova types using their light curves as inputs. Preprocessing these is crucial step that significantly affects final quality. In this talk, we study application multilayer perceptron (MLP), bayesian neural network (BNN), and normalizing flows (NF) approximate observations for single curve. We use approximations inputs demonstrate proposed methods outperform state-of-the-art Gaussian processes applying Zwicky Transient Facility Bright Survey curves. MLP demonstrates similar speed increase. Normalizing Flows exceeds terms approximation well.