Tensorizing LSTMs to make them wider and deeper without adding parameters and with minimal extra compute costs.
read moreAttention and Augmented Recurrent Neural Networks
Overview (with references) of attention and several types of augmentation for RNNs.
read more
