[bionhoward]

How could this help us understand the difference between the learned parameters and their gradients? Can the gradients become one with the parameters a la exponential function?