Laplace-domain waveform inversion using various optimizers.

Most full waveform inversions use gradient-based optimization methods for rapid convergence of objective function. Deep neural networks also use various gradient-based optimizers such as Momentum optimizer, Nesterov accelerated gradient, Adagrad, RMSProp and Adam optimizer for rapid convergence of loss functions in training step. Based on this similarity, we applied neural network's optimizers to Laplace-domain full waveform inversion algorithm. We compared each optimizer's convergence and inversion results using a salt benchmarking model.