How to fit a step function?
显示 更早的评论
The problem is only the first parameter (p(1)) is fitted, but p1(2) = p0(2), unchanged! Why?
X = [-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0];
Y = [0.1 -0.15 0.05 0.0 -0.05 10.2 10.5 11.5 8.2 9.2 10.0 10.5 10.2 10.9 10.5 11.5];
plot(X,Y)
step = @(x)x > 0;
fun = @(p,x) p(1)*step(x-p(2))
p0 = [9,-0.5];
p1 = lsqcurvefit(fun,p0,X,Y);
Z = fun(p1,linspace(min(X),max(X)));
hold on
plot(linspace(min(X),max(X)),Z,'r')
1 个评论
It's not clear to me what the fit should look like or the resolution of the steps. For example,
X = [-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0];
Y = [0.1 -0.15 0.05 0.0 -0.05 10.2 10.5 11.5 8.2 9.2 10.0 10.5 10.2 10.9 10.5 11.5];
plot(X,Y, 'b-o')
xr = [X(1),repelem(X(2:end),1,2)]; % X is row vec
yr = [repelem(Y(1:end-1),1,2),Y(end)]; % Y is row vec
hold on
plot(xr, yr, 'r-','LineWidth',2)
回答（1 个）
Figen Ece Demirer
2021-4-13
0 个投票
X = [-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0];
Y = [0.1 -0.15 0.05 0.0 -0.05 10.2 10.5 11.5 8.2 9.2 10.0 10.5 10.2 10.9 10.5 11.5];
cut=0;
plot(X,Y,'g--o')
step = @(x)x > cut;
step2 = @(x)x < cut;
fun = @(p,x) p(1)*step(x)+p(2)*step2(x);
p0 = [9,-0.5];
p1 = lsqcurvefit(fun,p0,X,Y);
Z = fun(p1,linspace(min(X),max(X)));
hold on
plot(linspace(min(X),max(X)),Z,'r')
1 个评论
Jörn Froböse
2021-6-10
lsqcurvefit works for continous functions only
类别
在 帮助中心 和 File Exchange 中查找有关 Least Squares 的更多信息
2021-6-10
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
