The autoencoding-RNN approach to irregular astronomical time series, at the LBNL machine-learning-for-science workshop.