Re: More on "memory exhausted..."

[Martin Helm]

On Dienstag, 2. August 2011 16:27:26 J Stasko wrote:
> Just to clear something up about my previous post...
> 
> My matrix is actually 217x217 or 200x200, not 47089x47089 or 40000x40000.
>  So I do not think my matrix size is the problem.  What I think is the
> problem is that fsolve is taking my 40000 or 47089 elements and squaring
> that to make a jacobian...  Does that make sense?
> 
> I am still trying to come up with a way to produce the Jacobian (yes,
> mathematically it's straightforward, but expressing it in a numerical way
> cleanly is another question.)
> 
> -John
Sorry that I misunderstood you first. I made a very simple example with fsolve 
with a 250x250 matrix which does not show your problem (octave 3.4.2 on 
openSUSE 11.4 64bit on my netbook with 3GB RAM). This is of course a plain 
stupid example from me

a = rand(250);
b=ones(250,1);
opt = optimset();
x = fsolve(@(x) a*x,b, opt);


I also watched the memory use to see if something strange happens, but it 
never went over 980 MB (this includes all the tasks from the desktop 
environment and other programs). So without looking into fsolve's source code 
this does not look like building a large matrix in memory while it runs.

Can you post your octave version, your operating system and architecture  and 
if possible a self contained example of code which triggers your problem? So 
that we can try to see if we can also reproduce the problem somehow?
Maybe you simply call fsolve in the wrong way?