Lipschitz optimization methods for fitting a sum of damped sinusoids to a series of observations

A general nonlinear regression model is considered in the form of fitting a sum of damped sinusoids to a series of non-uniform observations. The problem of parameter estimation in this model is important in many applications like signal processing. The corresponding continuous optimization problem is typically difficult due to the high multiextremal character of the objective function. It is shown how Lipschitz-based deterministic methods can be well-suited for studying these challenging global optimization problems, when a limited computational budget is given and some guarantee of the found solution is required.