1 paper · 1 filter
Guoqiang Zhang, Richard Heusdens
In this paper, we extend the standard Attention in transformer by exploiting the consensus discrepancy from a distributed optimization perspective, referred to as AttentionX. It is…