How Smoothing is N-simplicial Attention?

Alexandre Dussolle, Pietro Liò

Going from pure Multilayer Perceptron (MLP) to a learnable graph message-passing mechanism at each layer has been foundational to state-of-the-art results, despite the computationa…