Asked by sarah
on 11 Mar 2011

I need to write a function MySolve that will solve the nonlinear system of equations in the form

0=f(x)

where I have f:R^n -> R^n is an arbitrary function with n-dimensions input and n-dimensional output. The function should implement the newton iteration.

I know the first line of my function looks like:

function [x,converged]=MySolve(f,xo,tol,maxit)

But I am having difficulty with the rest

Answer by Matt Tearle
on 11 Mar 2011

function [x,converged]=MySolve(f,xo,tol,maxit) opts = optimset('MaxIter',maxit,'TolFun',tol); [x,~,converged] = fsolve(f,xo,opts);

(Unless, of course, this is a homework problem. In which case you might not be allowed to do that.)

Sean de Wolski
on 11 Mar 2011

Excellent Matt!

Andrew Newell
on 11 Mar 2011

@Matt, I'm afraid I'll have to dock you some marks for using the the trust-region dogleg algorithm instead of a Newton method.

Matt Tearle
on 11 Mar 2011

It's a fair cop. I was too lazy to look up which method fsolve used by default. From memory, I thought it was Levenburg-Marquardt. Which would count as a Newton method. Maybe it's fzero that uses Lev-Marq.

Answer by Sean de Wolski
on 11 Mar 2011

This should help a little

while the_error > tol && iter <= maxit iter = iter+1; Do a whole bunch of other stuff that you have to write end

Answer by Joe Howes
on 10 May 2012

i am also having the same problem as this so far i have

function [x,converged]=MySolve(f,xold,tol,maxit)

%maxit maximum number of iterations to be tried.

x=xold;

h=1e-10;

% run a loop from 1 to maxit

for k=0:maxit

%Need to call in MyJacobian

J=Myjacobian(f,x,h);

% Newton iteration

x= xold-(J\f(xold));

if(max(abs(x-xold)))<tol && (max(abs(x-xold)))<tol converged=1;

% Newtons iteration has converged

else converged=0;

%Newtons iteration hasn't converged

end

xold=x;

end

end

with my jacobian function looking like this

function df=Myjacobian(f,x,h)

% f: function to be differentiated %x: point where jacobian is taken % h: parameter for finite differences % outputs df: m*n matrix, jacobian of f in x.

n=length(x);

% defines number of rows

fx=f(x); m=length(fx);

% defines number of colums

df=zeros(n,m);

% matrix of zeros

for i=1:n;

% runs a loop that takes two values x1 and x2 and places it % into my empty matrix df the process then produces 2 matircies of % df1 and df2

x(i)=x(i)+h; x1= f(x); x(i)=x(i)-(2*h); x2= f(x); df(:,i)=(x1-x2)/(2*h); x(i)=x(i)+h;

end

end

Walter Roberson
on 10 May 2012

Do you want discussion here or in your Question on this topic, http://www.mathworks.com/matlabcentral/answers/38031-mysolve-help

Geoff
on 10 May 2012

Haha, yeah just gave a bunch of suggestions on that. Hopefully not too much =)

Just direct the whole class to this site. Your tutor would be so impressed.

Related Content

MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn moreOpportunities for recent engineering grads.

Apply Today
## 1 Comment

## Andrew Newell (view profile)

Direct link to this comment:http://www.mathworks.com/matlabcentral/answers/3007#comment_6276

No one is going to do your homework for you. Try solving this problem yourself first, then show us your code and tell us what problems you're having.