Title: GLU Variants Improve Transformer
Authors: Noam Shazeer
Published: 12th February 2020 (Wednesday) @ 19:57:13
Link: http://arxiv.org/abs/2002.05202v1
Abstract
Gated Linear Units (arXiv:1612.08083) consist of the component-wise product of two linear projections, one of which is first passed through a sigmoid function. Variations on GLU are possible, using different nonlinear (or even linear) functions in place of sigmoid. We test these variants in the feed-forward sublayers of the Transformer (arXiv:1706.03762) sequence-to-sequence model, and find that some of them yield quality improvements over the typically-used ReLU or GELU activations.