Introduction - If you have any usage issues, please Google them yourself
In the beginning of adaption, the big
error makes the big step size and fast convergence rate with
the adaptive process going on, errors are reduced gradually,
step size is also decreased. When adaption reached
steady-state phase, the error is very small, and close to zero,
step size is zero approximately