非线性优化方法在脑电逆问题中的应用与比较

Application and Comparison of Some Nonlinear Optimization Methods in EEG Inverse Problem

摘要: 讨论了基于单纯形法（Simplex）、阻尼最小二乘法（LM）方法、混合遗传算法（HGA）等非线性优化方法的脑电偶极子定位问题．仿真结果表明，这些算法在一定条件下均可使用，比较而言，Simplex、LM法有较快的计算速度，而HGA耗时较多．另一方面，HGA对迭代初值要求不高，而Simplex、LM则对此有一定的要求．特别在对多偶极子源求解时，Simplex、LM有时因初值选择困难而失效，而HGA仍然是有效的．

Abstract: This paper discusses EEG dipole source localization problems solved by nonlinear optimization methods, such as Simplex, Levenberg-Marquart and Hybrid Genetic Algorithm. Computer simulation demonstrates that these algorithms are all available under certain conditions. In comparison, Simplex, LM algorithms are of fast computation speed while HGA costs more time. On the other hand, HGA almost has no special need for the selection of initial iterative values while Simplex and LM algorithms haves such need to some extent. Furthermore, on estimation of multiple source parameters, Simplex, LM algorithms sometimes fail due to incorrect selection of initial iterative values, while HGA is still effective.