Asynchronous distributed solution of large scale nonlinear inversion problems (Q1294576)

The author considers large scale nonlinear least squares problems, the object functions of which are difficult to evaluate and the Jacobians of which are ill posed and have an almost diagonal structure. The author proves the convergence of an asynchronous iteration, which means that the components of the approximate vectors are asynchronous under certain assumptions on the nonlinear mapping operator being a vector contracting one. The method is naturally parallelizable, thus it can be used in practical inversion problems, such as in seismic exploration for hydrocarbons.