Neural inversion of Gastom´90 tomographic data

Ocean acoustic tomography is a synoptic observation method of the internal structure of the ocean. The goal of this paper is to present the results of a nonlinear method for tomography inversion in an acoustic environment with unresolved rays. This method relies upon the ability of neural nets to learn and generalize from examples. A set of sound velocity environments is built and the arrival time patterns in given experimental conditions are computed by a ray tracing model. In the first step, the inverse mapping is learned from this set of examples by a multilayered perceptrons. In a second step, when an unlearned time pattern is presented, the neural net estimates the unknown sound speed environment. Some significative results dealing with simulated data in the North-Atlantic environment are given and discussed