Accelerating nonlinear inversion algorithms on GPU platform for electromagnetic data

Nonlinear inversion algorithms have many applications in geophysical explorations, biomedical imaging, nondestructive testing and etc. One of the bottlenecks in these algorithms is the computational efficiency. In this abstract, we investigate accelerating the contrast source inversion algorithm using graphic process unit. By taking advantage of the massively parallel computing architecture of GPU, we can significantly reduce the computational time of inversion. Numerical examples show that we can reduce the computational time from hours to only a few minutes.