NDcost
generic external for optim computing gradient using finite differences
Syntax
[f,g,ind]=NDcost(x,ind,fun,varargin)
Arguments
- x
real vector or matrix
- ind
integer parameter (see optim)
- fun
Scilab function with syntax
F=fun(x,varargin)
varargin may be use to pass parametersp1,...pn
- f
criterion value at point
x
(see optim)- g
gradient value at point
x
(see optim)
Description
This function can be used as an external for
optim
to minimize problem where gradient is too
complicated to be programmed. only the function fun
which computes the criterion is required.
This function should be used as follow:
[f,xopt,gopt]=optim(list(NDcost,fun,p1,...pn),x0,...)
Examples
// example #1 (a simple one) //function to minimize function f=rosenbrock(x, varargin) p=varargin(1) f=1+sum( p*(x(2:$)-x(1:$-1).^2).^2 + (1-x(2:$)).^2) endfunction x0=[1;2;3;4]; [fopt,xopt,gopt]=optim(list(NDcost,rosenbrock,200),x0) // example #2: This example (by Rainer von Seggern) shows a quick (*) way to // identify the parameters of a linear differential equation with // the help of scilab. // The model is a simple damped (linear) oscillator: // // x''(t) + c x'(t) + k x(t) = 0 , // // and we write it as a system of two differential equations of first // order with y(1) = x, and y(2) = x': // // dy1/dt = y(2) // dy2/dt = -c*y(2) -k*y(1). // // We suppose to have m measurements of x (that is y(1)) at different times // t_obs(1), ..., t_obs(m) called x_obs(1), ..., x_obs(m) (in this example // these measurements will be simulated), and we want to identify the parameters // c and k by minimizing the sum of squared errors between x_obs and y1(t_obs,p). // // (*) This method is not the most efficient but it is easy to implement. // function dy=DEQ(t, y, p) // The rhs of our first order differential equation system. c =p(1);k=p(2) dy=[y(2);-c*y(2)-k*y(1)] endfunction function y=uN(p, t, t0, y0) // Numerical solution obtained with ode. (In this linear case an exact analytic // solution can easily be found, but ode would also work for "any" system.) // Note: the ode output must be an approximation of the solution at // times given in the vector t=[t(1),...,t($)] y = ode(y0,t0,t,list(DEQ,p)) endfunction function r=cost_func(p, t_obs, x_obs, t0, y0) // This is the function to be minimized, that is the sum of the squared // errors between what gives the model and the measurements. sol = uN(p, t_obs, t0, y0) e = sol(1,:) - x_obs r = sum(e.*e) endfunction // Data y0 = [10;0]; t0 = 0; // Initial conditions y0 for initial time t0. T = 30; // Final time for the measurements. // Here we simulate experimental data, (from which the parameters // should be identified). pe = [0.2;3]; // Exact parameters m = 80; t_obs = linspace(t0+2,T,m); // Observation times // Noise: each measurement is supposed to have a (gaussian) random error // of mean 0 and std deviation proportional to the magnitude // of the value (sigma*|x_exact(t_obs(i))|). sigma = 0.1; y_exact = uN(pe, t_obs, t0, y0); x_obs = y_exact(1,:) + grand(1,m,"nor",0, sigma).*abs(y_exact(1,:)); // Initial guess parameters p0 = [0.5 ; 5]; // The value of the cost function before optimization: cost0 = cost_func(p0, t_obs, x_obs, t0, y0); mprintf("\n\r The value of the cost function before optimization = %g \n\r",... cost0); // Solution with optim [costopt,popt]=optim(list(NDcost,cost_func, t_obs, x_obs, t0, y0),p0,... 'ar',40,40,1e-3); mprintf("\n\r The value of the cost function after optimization = %g",costopt) mprintf("\n\r The identified values of the parameters: c = %g, k = %g \n\r",... popt(1),popt(2)) // A small plot: t = linspace(0,T,400); y = uN(popt, t, t0, y0); clf(); plot2d(t',y(1,:)',style=5) plot2d(t_obs',x_obs(1,:)',style=-5) legend(["model","measurements"]); xtitle("Least square fit to identify ode parameters")
See also
- optim — non-linear optimization routine
- external — Scilab Object, external function or routine
- numderivative — approximate derivatives of a function (Jacobian or Hessian)
Report an issue | ||
<< Semidefinite Programming | Optimization and Simulation | fsolve >> |