rawlsq.m

Incorporating Prior Knowledge in Cubic Spline Approximation - Application to the Identification of R

function [P1,P2]  = rawlsq(cold,tt,cv,reler,ter,knots,cycnum)
%
% cold -> valodi gorbe
% tt ->  idovektor a 12 v. 31 ponthoz
% cc -> 12x3 v. 31x3 matrix adatokhoz
% reler -> rel.zaj 0.05 0.10
% ter -> zaj tipusa: 0=fuggetlen,1=fuggo
% knots -> csomopontok
% cycnum -> ciklusszam
% P1 <- p1 vector
% P2 <- p2 vector
%

seeki = 0; randn('seed',seeki);
P1 = []; P2 = [];


for II = 1:cycnum,
 %+noise
 ca = cv(:,1);
 cb = cv(:,2);
 cc = cv(:,3);
 if ter==0,
  ca = ca + randn(size(ca)).*ca*reler;
  cb = cb + randn(size(cb)).*cb*reler;
  cc = cc + randn(size(cc)).*cc*reler;
 else
  rr = randn(size(ca));
  ca = ca + rr.*ca*reler;
  cb = cb + rr.*cb*reler;
  cc = cc + rr.*cc*reler;
 end
 %P1
 t = tt;
 c = [ca cb cc];
 libarycommand = 1;
 rate_libary;
 P1 = [P1, norm((b'-[1 0.5 10 5])./[1 0.5 10 5])];
end