## Discretize B ## # [B,N] [E->N] [B,E] B = layer.W_B(x[b,l]) # no bias

Shouldn't this be x[:,l] instead of x[b,l]?

y_t=[N]C[E,N]h_t+[E]    <this one> E    [E]xt

Shouldn't this be 'D'?

E

Did you mean 'D' here? (2nd equation of the structured SSM)

Is this a typo? (Δtvt+1)xt−1