Analysis of the Regularization Parameter for EEG Inverse Problem Using Genetic Algorithm

Electroencephalography (EEG) inverse problem intends to determine the internal brain signals from EEG electrode signals. Being an ill-posed problem, it needs a regularization function and a regularization parameter (λ). Many techniques have been proposed with different regularization functions, yet all of them still depends on the regularization parameter that can drastically change the solution given its value. In this study we intend to determine if the regularization parameter can be set to a specific value or a smaller interval, and if not, discover what changes in the problem settings causes it to vary. For that we simulated source signals and calculated the signals that these would generate in the EEG electrodes. We then used a genetic algorithm to find the best value of λ applied in different experiments having different regularization functions. Further study is still required, but we could conclude that, for the settings used in this study, different techniques demand different values of λ, and for some experiments with different source signals the value of λ maintained in a smaller interval.