Supernovae classifies supernovae using their light curves directly as inputs to a deep recurrent neural network, which learns information from the sequence of observations. Observational time and filter fluxes are used as inputs; since the inputs are agnostic, additional data such as host galaxy information can also be included.