2000 character limit reached
Generalized Attention Mechanism and Relative Position for Transformer (2208.10247v1)
Published 24 Jul 2022 in cs.CL and cs.LG
Abstract: In this paper, we propose generalized attention mechanism (GAM) by first suggesting a new interpretation for self-attention mechanism of Vaswani et al. . Following the interpretation, we provide description for different variants of attention mechanism which together form GAM. Further, we propose a new relative position representation within the framework of GAM. This representation can be easily utilized for cases in which elements next to each other in input sequence can be at random locations in actual dataset/corpus.