When recurrent models don’t need to be recurrent

By John Miller An earlier version of this post was published on Off the Convex Path. It is reposted here with the author’s permission. In the last few years, deep learning practitioners have proposed a litany of different sequence models. Although recurrent neural networks were…