Skip to Main Content
The estimation of dipole parameters of EEG data is a difficult nonlinear optimization problem due to the multiple local minima in the cost function. In this paper, we present and evaluate a more robust and efficient optimization approach, named chaotic simulated annealing (CSA) algorithm. The key idea of CSA is to replace the Gaussian distribution by a chaotic sequence in the conventional standard simulated annealing (SSA). The effectiveness of the new method was confirmed by computer simulation and the preliminary application to the early event-related potential (ERP) of a spatial visual attention experiment. Our results showed that CSA has better robustness and higher feasibility compared with other methods for global optimization problems. Also, the positive results of dipole localization using CSA were achieved for an ERP study.