KilonovaNet: Surrogate Models of Kilonova Spectra with Conditional Variational Autoencoders

TL;DR

KilonovaNet employs conditional variational autoencoders to create fast, accurate surrogate models of kilonova spectra, significantly improving parameter inference efficiency in multimessenger astrophysics.

Contribution

This work introduces KilonovaNet, a novel cVAE-based surrogate modeling approach that directly learns from spectral data, eliminating preprocessing overhead and accelerating inference.

Findings

Surrogate models achieve low error on multiple datasets

Synthetic light curves enable effective parameter inference

KilonovaNet significantly speeds up analysis of kilonova observations

Abstract

Detailed radiative transfer simulations of kilonova spectra play an essential role in multimessenger astrophysics. Using the simulation results in parameter inference studies requires building a surrogate model from the simulation outputs to use in algorithms requiring sampling. In this work, we present KilonovaNet, an implementation of conditional variational autoencoders (cVAEs) for the construction of surrogate models of kilonova spectra. This method can be trained on spectra directly, removing overhead time of pre-processing spectra, and greatly speeds up parameter inference time. We build surrogate models of three state-of-the-art kilonova simulation data sets and present in-depth surrogate error evaluation methods, which can in general be applied to any surrogate construction method. By creating synthetic photometric observations from the spectral surrogate, we perform parameter inference for the observed light curve data of GW170817 and compare the results with previous analyses. Given the speed with which KilonovaNet performs during parameter inference, it will serve as a useful tool in future gravitational wave observing runs to quickly analyze potential kilonova candidates