Re: Nonlinear Equation Solve

[Ben Abbott]

On Feb 14, 2012, at 6:00 PM, Thomas D. Dean wrote:

> On 02/14/12 12:22, Thomas D. Dean wrote:
> octave:68> [x, fval, info] =
> fminunc (@eqnsetf,y',
>    optimset("TolX",1e-18,"TolFun",1e-18,"ComplexEqn","on"));
> error: __doglegm__: operator \: nonconformant arguments (op1 is 15x15, op2 is 
> 225x1)
> error: called from:
> error:   /usr/local/share/octave/3.4.3/m/optimization/fminunc.m at line 386, 
> column 3
> error:   /usr/local/share/octave/3.4.3/m/optimization/fminunc.m at line 233, 
> column 9
> 
> I am cut-paste-copy-edit challenged.
> I used 1e-18.
> 
> I am using 3.4.3 on a 64-bit machine, overclocked to 4.2GHz and 16GiB RAM.  
> TolX should be much smaller, I think.
> 
> Tom Dean

Check what the value of eps is. Just type "eps" as Octave's command line. On my 
64 bit machine I get ...

        eps
        ans =  2.2204e-16

Note, that Octave is still using double precision variables.

Does the optimization run without error if "TolX" and "TolFun" = 10*eps ?

Ben