no code implementations • ICLR 2018 • Daniel Levy, Danlu Chan, Stefano Ermon
In this work, we present LSH Softmax, a method to perform sub-linear learning and inference of the softmax layer in the deep learning setting.
Language Modelling