SLiCsim: light curves for the ML era
摘要
SLiCsim is the world’s first generic GPU-accelerated auto-differentiable supernova light-curve simulator. It is partly a port to the PyTorch ecosystem of the sncosmo package and partly an attempt to re-build the concept of SN simulation from the ground up with a focus on performance through batched parallelism and with ML/SBI applications in mind. This chapter briefly traces the simulation of a transient survey, explaining SLiCSim’s components (and philosophy) along the way.