Take fitted data from fit object - matlab

Question

I use the following example taken from the site that describes fit. Returns an
object. Is it possible to take the data that refer to the fitted surface?
load franke
sf = fit([x, y],z,'poly23')
plot(sf,[x,y],z)
Thanks!

Answer 1

Here is a way to do it; but there is probably a cleaner one:
After getting your sf object, you can access its methods like so:
MethodName(sf)
See here for the list of available methods
So let's say you wish to plot the surface using a handle for the plot:
hPlot = plot(sf)
Then you can fetch the XData, YData and ZData using the handles like so:
X = get(hPlot,'XData')
Y = get(hPlot,'YData')
Z = get(hPlot,'ZData')
Might be a it cumbersome but it works. Note that you can also fetch the coefficients of the fitted surface like so:
coeffvalues(sf)
and the formula used to generate it:
formula(sf)
Therefore you could generate X, Y data and create Z values using meshgrid for example and you could then modify the surface as you wish.
EDIT Here is how you could create your own surface using the coefficients and the formula. Here I create an anonymous function with two input arguments (x and y) and use it to generate z values to plot. From the data obtained using plot(sf) I used x = 1:0.01:0.01:1 and y = 500:500:3000 but you could obviously change them.
I entered the formula manually in the function handle but there has to be a better way; I'm a bit in a hurry so I did not looked further into that but you could extract every element of the formula and multiply it by the right coefficient to automatically generate the formula.
Here is the whole code:
clear
clc
close all
load franke
sf = fit([x, y],z,'poly23')
c = coeffvalues(sf)
F = formula(sf)
%// Generate x and y values.
[x,y] = meshgrid(500:100:3000,0.01:.01:1);
%// There should be a better approach than manually entering the data haha.
%// Maybe use eval or feval.
MyFun = #(x,y) (c(1) + c(2)*x + c(3)*y +c(4)*x.^2 + c(5)*x.*y + c(6)*y.^2 + c(7)*(x.^2).*y + c(8)*x.*y.^2 + c(9)*y.^3);
%// Generate z data to create a surface
z = (MyFun(x,y));
figure
subplot(1,2,1)
plot(sf)
title('Plot using sf','FontSize',18)
subplot(1,2,2)
surf(x,y,z)
title('Plot using MyFun','FontSize',18)
Output:

Answer 2

[I know this is many years later, but I thought I'd add this for future visitors looking for answers]
There is a much simpler way to do this - use feval, or the implicit shortcuts to it by calling the fittype object itself. It did take me a little digging to find this when I needed it, it isn't particularly obvious.
From the feval MATLAB documentation:
Standard feval syntax:
y = feval(cfun,x)
z = feval(sfun,[x,y])
z = feval(sfun,x,y)
y = feval(ffun,coeff1,coeff2,...,x)
z = feval(ffun,coeff1,coeff2,...,x,y)
You can use feval to evaluate fits, but the following simpler syntax is recommended to evaluate these objects, instead of calling feval directly. You can treat fit objects as functions and call feval indirectly using the following syntax:
y = cfun(x) % cfit objects;
z = sfun(x,y) % sfit objects
z = sfun([x, y]) % sfit objects
y = ffun(coef1,coef2,...,x) % curve fittype objects;
z = ffun(coef1,coef2,...,x,y) % surface fittype objects;