Curve fitting scilab how to#

* exp ( - x ( 2 ) * tm ) g = endfunction // now we could call leastsq: // 1- the simplest call = leastsq ( list ( myfun, tm, ym, wm ), x0 ) // 2- we provide the Jacobian = leastsq ( list ( myfun, tm, ym, wm ), mydfun, x0 ) // a small graphic (before showing other calling features) tt = linspace ( 0, 1.1 * max ( tm ), 100 ) ' yy = yth ( tt, xopt ) scf ( ) plot ( tm, ym, ' kx ' ) plot ( tt, yy, ' b- ' ) legend ( ) xtitle ( ' a simple fit with leastsq ' ) // 3- how to get some information (we use iprint=1) = leastsq ( 1, list ( myfun, tm, ym, wm ), mydfun, x0 ) // 4- using the conjugate gradient (instead of quasi Newton) = leastsq ( 1, list ( myfun, tm, ym, wm ), mydfun, x0, ' gc ' ) // 5- how to provide bound constraints (not useful here !) xinf = xsup = // without Jacobian: = leastsq ( list ( myfun, tm, ym, wm ), ' b ', xinf, xsup, x0 ) // with Jacobian : = leastsq ( list ( myfun, tm, ym, wm ), mydfun, ' b ', xinf, xsup, x0 ) // 6- playing with some stopping parameters of the algorithm // (allows only 40 function calls, 8 iterations and set epsg=0.01, epsf=0.1) = leastsq ( 1, list ( myfun, tm, ym, wm ), mydfun, x0, ' ar ', 40, 8, 0.01, 0.1 )

* ( yth ( tm, x ) - ym ) endfunction function g = mydfun ( x, tm, ym, wm ) v = wm. This vector (possibly matrix) of same size asįunction y = yth ( t, x ) y = x ( 1 ) * exp ( - x ( 2 ) * t ) endfunction // we have the m measures (ti, yi): m = 10 tm = ' ym = ' // measure weights (here all equal to 1.) wm = ones ( m, 1 ) // and we want to find the parameters x such that the model fits the given // data in the least square sense: // // minimize f(x) = sum_i wm(i)^2 ( yth(tm(i),x) - ym(i) )^2 // initial parameters guess x0 = // in the first examples, we define the function fun and dfun // in scilab language function e = myfun ( x, tm, ym, wm ) e = wm. Sequence of optional parameters controlling the convergence of Integer, number of variables used to approximate the Hessian Most of these reports are written on the Scilab standard Iprint=0 nothing (except errors) is reported, Scalar argument used to set the trace mode. Stand for quasi-Newton (default), conjugate gradient or X0 (-%inf and +%inf may be used for dimension The lower bound binf then by the upper boundīounds are real vectors with same dimension than Introduced by the string keyword 'b' followed by Real vector (initial guess of the variable to beĪ scilab function or a string defining the Jacobian matrix ofįun (see more details in DESCRIPTION). A scilab function or a list defining a function from

Curve fitting scilab