1 paper · 1 filter
Shai Zucker, Xiong Wang, Fei Lu +1
We study the convergence rate of learning pairwise interactions in single-layer attention-style models, where tokens interact through a weight matrix and a nonlinear activation fun…