What does the following error mean? Undefined function 'diffjac' for input arguments of type 'function_handle'. Error in Heq_Nwt (line 23) df = diffjac(x1,f,f0)

My code is below. Not sure why I cannot use diffjac function. I also get an error for input arguments of type 'double' if I change my diffjac inputs to (x1,f,f0).

function [ x, ex] = Heq_Nwt( f, x1, tol, nmax )

c=.8;
N=10;

x2 = ones(N,1);
x1 = transpose(x2);

meu=1:N;
meu=(meu-.5)/N;
meu=meu';
cs=.5*c/N;
ChH= meu';

in = @(x1) sum((ChH'*x1)./(bsxfun(@plus, ChH',ChH)));

n2=ones(N,1)';

f = @(x1)x1 - (n2-(cs*(in(x1)))).^-1
f0 = f(x1)

df = diffjac(x1,f,f0)

Answer 1

[Matthew]

Hi Branden,

What are the results of

which diffjac -all

?

If its empty, then you need to add the folder diffjac is in to your path. If its not, then diffjac is written for a different type of object, and to use it you will need to overload it and define your desired behavior from it.

Answer 2

[Branden Chamness]

Results are:

'diffjac' not found.

I added diffjac.m file to directory. It is as follows:

function [l, u] =diffjac(x, f, f0)
% compute a forward difference Jacobian f'(x), return lu factors
%
% uses dirder.m to compute the columns
%
% C. T. Kelley, November 25, 1993
%
% This code comes with no guarantee or warranty of any kind.
%
%
% inputs:
%         x, f = point and function
%      f0   = f(x), preevaluated
%
n=length(x);
for j=1:n
    zz=zeros(n,1);
    zz(j)=1;
    jac(:,j)=dirder(x,zz,f,f0);
end
[l, u] = lu(jac);

I received the following errors when running original code once diffjac.m added:

Undefined function 'dirder' for input arguments of type 'function_handle'.

Error in diffjac (line 19)
    jac(:,j)=dirder(x,zz,f,f0);

Seems I need to add this file to directory as well. So I did. It is as follows:

function z = dirder(x,w,f,f0)
% Finite difference directional derivative
% Approximate f'(x) w
% 
% C. T. Kelley, November 25, 1993
%
% This code comes with no guarantee or warranty of any kind.
%
% function z = dirder(x,w,f,f0)
%
% inputs:
%           x, w = point and direction
%           f = function
%           f0 = f(x), in nonlinear iterations
%                f(x) has usually been computed
%                before the call to dirder

%
% Hardwired difference increment.
epsnew=1.d-7;
%
n=length(x);
%
% scale the step
%
if norm(w) == 0
    z=zeros(n,1);
return
end
epsnew = epsnew/norm(w);
if norm(x) > 0
    epsnew=epsnew*norm(x);
end
%
% del and f1 could share the same space if storage
% is more important than clarity
%
del=x+epsnew*w;
f1=feval(f,del);
z = (f1 - f0)/epsnew;

Now I am receiving errors from this file as well:

Error using  + 
Matrix dimensions must agree.

Error in dirder (line 38)
del=x+epsnew*w;

Am I going to have to debug diffjac.m and dirder.m in order to use diffjac in my code?

Answer 3

[Ahmet Cecen]

diffjac.m and dirder.m works perfectly fine for me. Your inputs must be different than intended.

Answer 4

[Branden Chamness]

Should I match my inputs to match diffjac and dirder? For example my input is 'x1' while diffjac and dirder is 'x'.

See any issue with my function? I used anonymous function. Is diffjac/dirder looking for different function notation?

f = @(x1)x1 - (n2-(cs*(in(x1)))).^-1