Standard Residual vs Attention Residual

Standard Residual
Every layer gets weight 1
Attention Residual
Learned, input-dependent weights
Press Play to see dynamic weight selection