What Dense Graph Do You Need for Self-Attention?. (arXiv:2205.14014v4 [cs.LG] UPDATED)

Web: http://arxiv.org/abs/2205.14014

Transformers have made progress in miscellaneous tasks, but suffer from
quadratic computational and memory complexities. Recent works propose sparse
Transformers with attention on sparse graphs to reduce complexity and remain
strong performance. While effective, the crucial parts of how dense a graph
needs to be to perform well are not fully explored. In this paper, we propose
Normalized Information Payload (NIP), a graph scoring function measuring
information transfer on graph, which provides an analysis tool for trade-offs
between performance and …

arxiv attention graph lg self-attention