Leastsq function for z=f(x,y) problem

Specific domains Optimization
1

Hello,
Is it possible to use leastsq function on a 3D problem like z=f(x,y)?
if yes, can anyone share an example?
BR

My best guess so far

clc
clear
alvclose(winsid())

// Function definition
function F=funct(sol,X,Y)
F=1-exp(-(sol(1).*Y+sol(2)).*X.^2)+grand(size(Y,‘r’),size(Y,‘c’),“nor”,0,0.025)
// added noise
endfunction

// Function parameters
tmax=3
ao=4e-2
bo=1e-2
time=linspace(0,tmax,100)’
temp=linspace(20,80,4)’
[X,Y]=meshgrid(time,temp)
Z=funct([ao bo],X,Y)

// plot the noisy data
scatter3d(X,Y,Z,‘.’)
gca().zoom_box=[0 0 3 100 -0.2 1.2]

// Error definition
function dist = Error(sol, X, Y, Z)
dist = Z - funct(sol, X, Y);
endfunction

eps=10*%eps;
level=1000
p0 = [1 1]‘*1e-2;
[xopt, fopt, info] = leastsq(list(Error, X, Y, Z), p0’)//, [eps,eps,eps,level,eps,level]);

2

Hello,

You just have to reshape the residual matrix dist, i.e.

dist=dist(:)

so that leastsq sees a classical vector residual.

S.

3

Of course … I’m kidding … you remove a stone from my shoe
Thank you
F