Sampled Transformer for Point Sets

Shidi Li, Christian Walder, Alexander Soen +2

The sparse transformer can reduce the computational complexity of the self-attention layers to $O(n)$, whilst still being a universal approximator of continuous sequence-to-sequenc…