1 paper
Xiaojun Tan, Yuchen Zhao
The statistical essence of the Transformer architecture has long remained elusive: Is it a universal approximator, or a neural network version of known computational algorithms? Th…