Comment by badmonster

How does BDH handle long-range dependencies compared to Transformers, given its locally interacting neuron particles? Does the scale-free topology implicitly support efficient global information propagation?