Numerical inverse results shows that the gradient regularization method has little dependence on the prior information of problem solution, which solves the sensitivity of the initial values to inv...