In general inverse problems, given a forward process y=f(x), the goal is to find a suitable inverse model x=f^(-1) (y) to map the inverse process where y is the experimental measurement. f is physical law and can be implemented as a forward simulation but f^(-1) is difficult to solve analytically and usually viewed as optimization problems, in which x are modified iteratively, such that the predictions from forward simulation match the measurements. However, this optimization problem is iterative and, in some cases, very computationally expensive, especially when ∂f⁄∂x is not possible to obtain or intractable.