Hybrid Super Memory Gradient Method Full Waveform Inversion

In this paper, we discussed the Hybrid Super Memory Gradient (HSPM) method for FWI, whereby the Super Memory Gradient (SPM) method and Unconstrained Super Memory Gradient (USPM) method are used for FWI iterations. SPM method uses many gradients information to correct current descent direction, so we can obtain more accurate descent direction. USPM method without Wolfe linear search which can save a lot of computational time. Firstly, at low frequencies, we use USPM method , and then use SPM method to inverse high frequencies information. The numerical examples show that HSPM delivers similar results as those expected with the SPM method, and improved computational efficiency over Conjugate Gradient method.