Yamaoka K, Nakagawa T
J Pharmacobiodyn. 1983 Aug;6(8):595-606. doi: 10.1248/bpb1978.6.595.
A nonlinear least squares program based on first-order simultaneous differential equations, MULTI (RUNGE), for personal computers was developed. Four algorithms, i.e. the classical Gauss-Newton method, the damping Gauss-Newton method, the modified Marquardt method and the simplex method can be used for the nonlinear curve fitting in MULTI (RUNGE). The differential equations and the initial conditions which are voluntarily defined by the user are numerically solved by the Runge-Kutta-Gill method and the numerically integrated solutions are fitted to the observed data points by the iterative least squares algorithms. MULTI (RUNGE) can be executed on many personal computers without any modification, because this program is written in the Microsoft minimum BASIC commands alone. The dimensions of differential equations, the number of parameters to evaluate and the number of experimental data points are restricted only by the available memory in computers and by the computing time.
开发了一种基于一阶联立微分方程的非线性最小二乘程序MULTI (RUNGE),用于个人计算机。四种算法,即经典高斯-牛顿法、阻尼高斯-牛顿法、修正的马夸特法和单纯形法,可用于MULTI (RUNGE)中的非线性曲线拟合。用户自行定义的微分方程和初始条件由龙格-库塔-吉尔方法进行数值求解,数值积分解通过迭代最小二乘算法拟合到观测数据点。MULTI (RUNGE)可以在许多个人计算机上直接运行,无需任何修改,因为该程序仅用微软最小BASIC命令编写。微分方程的维数、待评估参数的数量和实验数据点的数量仅受计算机可用内存和计算时间的限制。