Talks

Stable, Fast and Accurate: Kernelized Attention with Relative Positional Encoding

Your Transformer May Not be as Powerful as You Expect

Is $L^2$ Physics-Informed Loss Always Suitable for Training Physics-Informed Neural Network?