Abstract: In general, geophysical inversion is a nonlinear, multi-parameter, non-onvex question. It is therefore worthy of solving geophysical inversion questions with nonlinear approaches to search for new global optimization algorithm. In the present work, the thought of chaotic noise perturbing is introduced into
seismic waveform inversion. The method is characterized by introducing a simple nonlinear feedback into the gradient descendant systems for optimization tasks. It not only maintains the fast searching speed of the gradient descendant method, but also is of the capability to get away from the pitfall of local extremum. Results show that in comparison with conventional genetic algorithm and logistic map-based chaotic optimization algorithm the proposed method is faster in computation and needs
less subjective parameters, and then is easier to lead to a global optimization.

Key words: chaos, gradient descendant, optimization algorithm, seismic inversion