Creating the NN architecture means coming up with values for the number of layers of each type and the number of nodes in each of these layers.
Basic points for consideration in RNNs:
Mathematically speaking, frequentist and Bayesian methods differ in what they care about, and the kind of errors they’re willing to accept.
Questions that I need answers to
Training recurrent neural networks with Ilya Sutskever (PhD thesis): http://www.cs.utoronto.ca/~ilya/pubs/ilya_sutskever_phd_thesis.pdf
In this article we discuss basic differences between Generative and Discriminative models. This article is still in its draft mode and should be taken as such.