variational recurrent neural networks (vrnn)

A method of introducing stochasticity into RNNs by using variational inference, which can improve their ability to model complex data sequences.