The regression model is y = f(X,θ) + ε . The vector y contains potentials measured on N sensors. f(X,θ) is the vector with modeled potentials. X is the matrix with sensor coordinates, θ the vector with P source parameters of D sources, and ε the noise vector. Let G be the N by P matrix with the first order partial derivatives of the N modeled potentials to the P parameters. The covariance matri...